Practical, in-depth breakdowns of the strategies shaping modern search — written from the trenches of managing 5M+ monthly organic visitors at BankBazaar.
Click any article to read the full piece. More articles added regularly across Technical SEO, AEO, GEO, Programmatic SEO, and AI-driven growth.
AI-powered search is answering questions without clicks. Here's the full playbook to make your content the source Google AI Overviews, ChatGPT, Perplexity, and Gemini actually cite.
Answer Engine Optimisation (AEO) is the practice of structuring and presenting your content so that AI-powered search systems — Google AI Overviews, Bing Copilot, ChatGPT, Perplexity, Gemini, and voice assistants — select it as the direct answer to a user's question.
Where traditional SEO focuses on earning a ranked position on a search results page, AEO focuses on earning the zero position — the single answer that gets read aloud, surfaced in an AI summary, or displayed as a featured snippet before any organic results.
Think of it this way: traditional SEO gets you onto the shelf in a library. AEO makes you the book the librarian hands directly to the person asking a question.
AEO is not a replacement for SEO — it's an evolution of it. The technical foundations (crawlability, page speed, authority) still matter. AEO layers structured, authoritative, conversational content on top of a solid SEO base.
The search landscape has fundamentally changed. Users no longer just search — they ask. "What is the best credit card for travel?" is a question, not a keyword. And AI systems are increasingly built to answer questions, not just return a list of pages.
Here's what's driving this shift:
At BankBazaar, our shift toward AEO — restructuring key pages with FAQ schema, conversational Q&A sections, and entity-based content — resulted in BankBazaar being cited in Google AI Overviews, ChatGPT, and Perplexity for high-intent financial queries. This is the real-world impact of AEO done right.
To optimise for answer engines, you need to understand how they select answers. The process broadly involves three stages:
The AI system retrieves candidate pages from its index (or the live web, in the case of Perplexity and AI Overviews). Pages need to be crawlable, indexed, and authoritative. This is where traditional SEO still matters — if Google can't crawl your page, no AI system will surface it.
The AI parses the retrieved content and identifies whether it directly, clearly answers the question at hand. Content that directly states a definition, explanation, or step-by-step process — ideally in the opening 100 words — wins here. Vague, verbose introductions lose.
AI systems cross-reference signals of authority and trust: backlink profiles, brand mentions across the web, E-E-A-T signals, schema markup, and author credentials. A technically perfect answer from an untrustworthy domain won't be selected.
Structure every important piece of content with a "direct answer first" approach: state the core answer in the opening paragraph, then expand. Google's AI systems and featured snippet algorithms both reward this pattern.
| Dimension | Traditional SEO | AEO |
|---|---|---|
| Target | Ranked position on SERP | Zero position / direct AI citation |
| Query type | Keywords (head terms) | Questions & conversational queries |
| Content format | Long-form articles, landing pages | Structured Q&A, definitions, step lists |
| Success metric | Rankings, organic traffic | AI citations, featured snippets, voice answers |
| Schema focus | Breadcrumbs, products | FAQPage, HowTo, Speakable, QAPage |
| User intent | Broad intent signals | Specific question intent matching |
| Authority signals | Backlinks, DA | E-E-A-T, brand mentions, author expertise |
Based on research, experimentation at scale, and real-world results across BankBazaar's 15K+ page site, these are the signals that most consistently predict AEO success:
Schema markup is the single highest-leverage technical intervention for AEO. It communicates directly to AI systems exactly what your content is about, how it's structured, and what questions it answers.
The most impactful schema type for AEO. Mark up every significant Q&A on your page. At BankBazaar, implementing FAQPage schema across credit card and loan pages drove a measurable increase in AI Overview appearances within 60 days.
For any process-based content — "how to apply for a personal loan", "how to improve CIBIL score" — HowTo schema with numbered steps is a direct signal to AI systems that your content provides structured procedural guidance.
Marks specific text sections as suitable for text-to-speech delivery. Critically important for voice search optimisation and increasingly used by Google Assistant to select spoken answers.
Signals authorship, publication date, and editorial context. Essential for E-E-A-T. Combined with sameAs links to the author's LinkedIn or Google Scholar profile, it builds a strong author entity.
Different from FAQPage — used when a page presents a community Q&A format (like a forum or support page). Allows AI systems to identify both the accepted answer and the question in context.
Don't mark up content in schema that doesn't appear visibly on the page. Google explicitly penalises hidden or mismatched structured data. Every FAQ in your JSON-LD must have a corresponding visible question and answer on the page.
The way you structure and write content directly affects whether AI systems choose to cite it. Here is the framework I use across BankBazaar's high-traffic pages:
Replace generic headings like "Benefits of Credit Cards" with question-led headings like "What are the benefits of using a credit card?" This directly mirrors how users phrase queries in conversational search and voice search, making it far easier for AI systems to match your content to the right query.
Borrowed from journalism: lead with the most important information, then add detail. AI systems and featured snippet algorithms both scan the first 100 words of a section first. If the answer isn't there, they move to the next candidate page.
Google's E-E-A-T framework (Experience, Expertise, Authoritativeness, Trustworthiness) is the foundation of AEO credibility. AI systems don't just evaluate content — they evaluate the source of content.
Traditional SEO metrics don't fully capture AEO performance. Here are the metrics that matter:
| Metric | What It Measures | Where to Track |
|---|---|---|
| Featured Snippet appearances | Zero-position wins in Google | Google Search Console, SEMrush |
| AI Overview citations | Presence in Google's AI summaries | Manual checks, AI visibility tools |
| People Also Ask wins | Presence in PAA boxes | SEMrush, Ahrefs, manual SERP checks |
| Voice search answers | Content read by Google Assistant | Speakable schema reports |
| Brand mentions in AI tools | Citations in ChatGPT, Perplexity, Gemini | Manual prompting, brand monitoring tools |
| FAQ click impressions | FAQ schema activations in SERPs | Google Search Console rich results |
Ready to implement? Here's the prioritised action plan, ordered by impact-to-effort ratio:
AEO is not about gaming AI systems — it's about being genuinely, demonstrably helpful. The brands that will dominate AI-powered search are the ones that answer questions better, faster, and more credibly than anyone else. Build for the user, structure for the machine.
Want to discuss AEO strategy or SEO collaboration?
Get in Touch with Maheswari →ChatGPT, Gemini, and Perplexity are the new search engines for millions of users. If your brand isn't being cited by these AI systems, you don't exist in their world. Here's how to change that.
Generative Engine Optimisation (GEO) is the discipline of making your brand, content, and expertise visible and trustworthy to large language model (LLM)-powered AI systems — including ChatGPT, Google Gemini, Perplexity AI, Claude, Bing Copilot, and emerging AI assistants.
Unlike traditional SEO, which targets Google's algorithmic ranking of pages, GEO targets the training data, real-time retrieval, and citation decisions of generative AI models. The goal: when someone asks an AI assistant about your topic, industry, or use case, your brand is the one that gets mentioned, cited, and recommended.
GEO is sometimes called LLM SEO, AI search optimisation, or generative AI visibility — all describing the same emerging discipline.
If someone asks ChatGPT "What's the best credit card for cashback in India?" and BankBazaar is mentioned as a trusted source — that's GEO working. If a competitor is mentioned instead, that's a GEO gap to close.
Every few years, a paradigm shift forces SEOs to fundamentally rethink strategy. Mobilegeddon in 2015 made mobile-first non-negotiable. The BERT update in 2019 made semantic search understanding critical. GEO is the 2025–2026 equivalent — and it's bigger than either.
Here's why this shift is structural, not cyclical:
At BankBazaar, our investment in GEO-aligned content — structured data, entity optimisation, E-E-A-T, and authoritative off-page citations — has resulted in consistent mentions in ChatGPT, Perplexity, and Google Gemini for key financial verticals. This is measurable, trackable brand presence in AI-powered search.
Different AI systems work differently, and understanding the distinction is critical for GEO strategy.
Some AI responses are based primarily on pre-training data — the vast corpus of text that was used to train the model before its knowledge cutoff. For these responses, GEO is about ensuring your content was present and prominent in high-quality, widely-indexed web content that was included in training datasets. Wikipedia, major news publications, industry reports, and highly-linked blog content are typically over-represented.
Perplexity, Bing Copilot, and Google AI Overviews use real-time retrieval — they search the live web, retrieve relevant pages, and use those pages to generate their answer. For these systems, GEO heavily overlaps with AEO: your content must rank well enough to be retrieved, and must then be structured to be cited.
ChatGPT with browsing enabled, Google Gemini, and most modern AI assistants use a hybrid: the base LLM provides general knowledge, real-time retrieval supplements it for current queries. GEO must address both layers.
| Dimension | AEO | GEO |
|---|---|---|
| Target | Featured snippets, voice answers, AI Overviews | ChatGPT, Gemini, Perplexity citations & recommendations |
| Primary channel | Google Search (AI-enhanced) | Standalone AI tools and AI-native search engines |
| Key lever | Schema markup, question-led content | Entity authority, off-page brand mentions, training data presence |
| Measurement | Featured snippet wins, AI Overview appearances | AI tool citations, brand mentions in AI responses |
| Timeline | Weeks to months (retrieval-based) | Months to years (training-cycle dependent) |
| Overlap | High — both require E-E-A-T, structured content, and authority signals | |
Based on emerging research — including the Stanford/Georgia Tech GEO paper (2024) and extensive real-world testing — these are the signals most correlated with AI citation frequency:
GEO-aligned content is different from traditional SEO content in key ways. Here's the framework:
AI systems understand the world through entities — specific, identifiable things (BankBazaar, CIBIL score, RBI, HDFC Bank). Content that clearly references and explains relationships between relevant entities is more likely to be understood, indexed accurately, and cited by AI systems.
Instead of writing "the best credit cards offer good rewards", write "HDFC Bank's Infinia credit card and Axis Bank's Magnus card offer among the highest reward rates in India, according to BankBazaar's 2026 credit card comparison." This is entity-rich, specific, and citable.
AI systems — especially those trained on curated web corpora — disproportionately represent original research, proprietary data, and unique statistics. Publishing original surveys, proprietary data analyses, and first-party research dramatically increases your chances of being included in training data and real-time citations.
AI systems favour sources that demonstrate deep, comprehensive expertise on a topic rather than surface-level coverage. A site with 50 deeply interconnected articles on personal loans is more likely to be cited for personal loan queries than a site with one great article. Build content clusters, not isolated pages.
Write in clear, precise sentences that can stand alone as a cited quote. Avoid ambiguous pronoun references ("it", "they") and ensure every key claim is a complete, attributable sentence. AI systems extract and cite specific passages — make those passages clean and quotable.
Before publishing, ask: "If an AI system extracted a single sentence from this page to answer a query, would that sentence accurately and completely represent my brand's position?" If yes, the content is GEO-ready.
Entity optimisation is the practice of ensuring your brand, people, products, and expertise are clearly defined, consistently represented, and widely referenced across the web — so that AI systems build an accurate, rich knowledge graph entry for your entity.
A verified Google Knowledge Panel is one of the strongest signals that your brand is a real, established entity. Claim it via Google's Knowledge Panel claim process, link it to your official website, social profiles, and Wikipedia/Wikidata page.
Wikipedia is heavily over-represented in LLM training data. A well-maintained, neutral Wikipedia article about your brand or key people dramatically increases the probability of accurate LLM knowledge. Even a Wikidata entry with structured properties (founded, headquarters, industry, key people) adds significant entity clarity.
Your brand name, description, founding date, key people, products, and category must be consistent across LinkedIn, Crunchbase, industry directories, news databases, and social platforms. Inconsistent information confuses AI knowledge graphs.
Individual authors are entities too. Build your personal entity: LinkedIn profile, Google Scholar profile (if applicable), bio pages on publication websites, consistent use of your full professional name, and links between all these profiles. When an AI system attributes expertise on a topic to "Maheswari P, SEO expert at BankBazaar", that attribution is powered by entity clarity.
Off-page GEO is arguably more important than on-page — because LLMs learn about your brand primarily from what other sources say about you, not just from your own website.
Secure placements in top-tier publications: major newspapers, industry journals, government or regulatory reports, academic citations. Each placement in a high-authority source increases your probability of appearing in AI training datasets and real-time retrieval results.
Be present where users ask questions. Answer relevant questions on Reddit, Quora, and niche forums. Participate in LinkedIn discussions. Comment in industry threads. AI systems that use real-time retrieval pull heavily from these platforms — and your consistent, authoritative presence in community discussions builds a strong off-page entity footprint.
YouTube transcripts, podcast show notes, and webinar summaries are increasingly indexed and included in AI training data. Appearing as a guest expert on relevant podcasts or YouTube channels creates additional entity touchpoints across platforms.
Track every mention of your brand across the web (using tools like Brand24, Mention, or Google Alerts) and work to convert unlinked mentions into linked citations. Every additional inbound link from an authoritative source strengthens both your traditional SEO and your GEO entity graph.
ChatGPT's base knowledge comes from its training data. For real-time browsing (ChatGPT with search enabled), your content must rank highly enough to be retrieved. Prioritise: Wikipedia presence, widely-published original research, and strong backlink authority. ChatGPT browsing uses Bing — so Bing SEO is a GEO lever here.
Perplexity is retrieval-first — it searches the live web for every query. This means traditional SEO signals (rankings, crawlability, content quality) directly determine Perplexity citations. Additionally, Perplexity's "Copilot" mode favours content that is clear, structured, and directly answers the stated question. AEO practices are your GEO strategy for Perplexity.
Google uses its existing search index for Gemini's retrieval layer. Strong Google SEO fundamentals, schema markup, E-E-A-T signals, and FAQPage schema all directly improve your probability of being cited. Google also weights brands with established Knowledge Panels and Merchant Center profiles.
Bing Copilot uses Microsoft's search index. Ensure your content is indexed by Bing (submit to Bing Webmaster Tools), has strong structured data, and appears in Bing's top results for your target queries.
GEO measurement is newer and less tool-supported than traditional SEO — but it's rapidly maturing. Here's how to track it today:
| GEO Metric | How to Measure | Frequency |
|---|---|---|
| AI tool brand mentions | Manually query ChatGPT, Gemini, Perplexity with target queries. Log citations. | Weekly |
| Perplexity citation rate | Query your top 20 keywords in Perplexity. Track if your domain is cited. | Weekly |
| AI Overview appearances | Track via SEMrush AI Overview tracking or manual SERP checks. | Monthly |
| Brand mention volume | Brand24, Mention, or Google Alerts for brand mentions across platforms. | Ongoing |
| Wikipedia traffic & edits | Wikipedia page analytics (Wikistats). | Monthly |
| Knowledge Panel presence | Manual check: search your brand name in Google. | Monthly |
| Branded search volume | Google Search Console branded query impressions & clicks. | Monthly |
Here is a prioritised 90-day GEO implementation roadmap, ordered by impact:
GEO is not a tactic — it's a long-term brand strategy. The brands that invest in being genuinely authoritative, widely mentioned, and consistently helpful across the web will compound their AI visibility over time. Start now, measure consistently, and build for the long game.
Interested in GEO strategy or working together on AI search visibility?
Connect with Maheswari →Nearly two thirds of all Google searches now end without a click. Here's how to measure success, adapt your content strategy, and win visibility even when users never reach your website.
64.82% of Google searches now end without a single click to any website. This is not a temporary disruption caused by AI — it's the acceleration of a trend that has been building since featured snippets emerged in 2014. What AI Overviews and Google's AI Mode have done is compress that timeline dramatically.
Here's the paradox: while overall CTR drops when AI Overviews appear, the brands that are actually cited in those overviews see a +35% CTR boost compared to non-cited brands on the same SERP. Zero-click doesn't mean zero opportunity — it means the opportunity has moved upstream, to being the source AI cites rather than the link users click.
Three platform changes collided in 2025–2026 to drive zero-click to its current level:
AI Overviews now appear in approximately 18–21% of all Google searches globally — heavily concentrated on informational and how-to queries, which historically drove the most organic traffic. When an AI Overview answers the question, the need to click a blue link is eliminated for most users.
Launched in March 2025 and rolled out globally through 2025–2026, AI Mode is a full conversational search experience within Google. Users can ask multi-part questions and receive synthesised answers drawing from multiple web sources. 93% of AI Mode sessions end without an external website visit — because users get what they need in the conversation.
These predate AI but continue to grow in scope. Knowledge panels, People Also Ask boxes, local packs, and shopping graphs all serve information without requiring clicks. AI has simply added a much more powerful layer on top of these existing zero-click surfaces.
Zero-click impact is not uniform. Informational and educational content ("how to," "what is," "why does") faces near-total cannibalisation from AI Overviews. Commercial and transactional queries — comparisons, pricing, provider-specific searches — retain much higher click rates. For fintech brands like BankBazaar, this means the blog traffic funnel is shrinking while comparison and product pages remain relatively protected.
The zero-click headline numbers are real, but averages obscure important nuances. Here's what the 2026 research cohort shows:
| Query Type | AI Overview Rate | Click Rate | Implication |
|---|---|---|---|
| "How to" queries | ~99.9% | Very low | Content mostly cannibalised; focus on being cited |
| "What is" / definitional | High | Low | Knowledge panel + EEAT authority matters |
| Comparison queries | Moderate | Moderate-High | Comparison tables + original data win clicks |
| Brand + product queries | Very low | High | Traditional SEO still fully applies |
| Transactional ("buy X") | 3–4% | Very High | PDP and category pages — least disrupted |
Despite the headline numbers, clicks haven't disappeared — they've concentrated. The remaining 35% of searches that do generate clicks are increasingly dominated by three user intent categories:
Users who already know they want to go to a specific website ("BankBazaar credit card comparison", "Zerodha login") will always click. Brand SEO, branded search volume, and ensuring your own site ranks #1 for your brand terms is non-negotiable.
Users who want to buy, apply, sign up, or compare products with pricing are far less satisfied by AI-generated summaries. They need to see the actual product, the real rate, the current offer. E-commerce, fintech product pages, and service booking pages retain strong click rates.
A significant portion of AI search users — especially for health, legal, and financial information — use AI to get an initial answer but then click through to verify. Pew Research confirms this: people use AI for exploration and synthesis, then use traditional search to fact-check. High-authority informational content still drives verification clicks even after zero-click AI answers.
Being cited in an AI Overview is now worth more than being ranked #3 on a page with an AI Overview. Prioritise AEO and GEO tactics for informational content: direct answer structure, FAQ schema, E-E-A-T signals, and original data that AI systems want to reference.
AI Overviews struggle to replace detailed comparison tables with real pricing, user reviews, and live data. For fintech brands, this means investing in dynamic comparison tools, rate tables, and product detail pages — the content that converts, not just the content that informs.
In a zero-click world, brand mentions in AI answers, featured snippets, and knowledge panels deliver marketing value even without a click. Track "AI share of voice" — how often your brand is mentioned across ChatGPT, Perplexity, and Google AI Mode — as a primary brand KPI alongside organic traffic.
AI systems can summarise generic content from thousands of sources. What they can't replace is your proprietary data — original surveys, internal benchmarks, platform-specific findings. Original research becomes the only truly zero-click-proof content investment.
Brands are 6.5× more likely to be cited by AI systems through third-party sources than from their own domain. Securing placements in high-authority publications, industry reports, and regulatory databases is now a direct lever for AI search visibility — not just a backlink strategy.
| Old Metric | New Metric | How to Track |
|---|---|---|
| Organic clicks | AI citation rate | Manual AI query monitoring; tools like AI Traffic Analytics |
| Organic impressions | AI share of voice | Track brand mentions across ChatGPT, Perplexity, Gemini |
| CTR | Citation + click blended rate | GSC + AI citation tracking combined |
| Position #1 | AI Overview inclusion | SEMrush AI Overview tracking / manual SERP audit |
| Organic revenue | Branded search + direct traffic growth | GA4 channel attribution |
Zero-click search doesn't mean zero value. It means the value your content delivers to users now happens inside the search engine — and your job is to be the source that delivers it. Measure accordingly: brand authority, AI citation rate, and branded search volume are the new primary KPIs for informational content in 2026.
Want to audit your zero-click exposure and build a citation-first content strategy?
Connect with Maheswari →GEO helps your brand get cited in AI search results. LLMO ensures your content becomes part of what AI models actually know. Here's the distinction — and why it matters more than most SEOs realise.
Large Language Model Optimisation (LLMO) is the practice of structuring content, building semantic authority, and establishing citation reliability such that large language models — ChatGPT, Claude, Gemini, Perplexity — accurately represent, reference, and recommend your brand during generation.
While GEO focuses on whether AI search retrieval systems select your content as a cited source at query time, LLMO addresses a deeper layer: whether your brand's knowledge, expertise, and positioning are baked into the model's parametric memory — the knowledge stored in its weights from training data — and whether your content is consistently retrievable and accurately citable when the model searches live at inference time.
GEO influences whether content is selected for AI answers. LLMO influences whether that content becomes part of what models can reference during generation — even when no live retrieval is performed. Think of GEO as your search ranking, and LLMO as your brand's reputation inside the model itself.
| Dimension | SEO | GEO | LLMO |
|---|---|---|---|
| Goal | Rank pages for clicks | Be cited in AI-generated answers | Be accurately known and trusted by AI models |
| Primary surface | Google SERPs | AI Overview, ChatGPT, Perplexity | LLM parametric memory + retrieval |
| Key signals | Backlinks, relevance, freshness | Structure, E-E-A-T, entity clarity | Semantic grounding, citation reliability, source consistency |
| Measurement | Rankings, clicks, impressions | Citation rate, AI share of voice | Brand accuracy in AI responses, training data presence |
| Time horizon | Weeks to months | Months | Long-term (training cycles) |
Modern LLMs operate in two modes simultaneously. Parametric memory is knowledge encoded into the model's weights during training — this is what the model "knows" without looking anything up. RAG (Retrieval Augmented Generation) is when the model searches the live web at inference time to supplement its base knowledge with current information.
ChatGPT, as of early 2026, only activates web search on about 34.5% of queries — down from 46% in late 2024. This means the majority of ChatGPT responses still rely entirely on parametric memory. If your brand isn't well-represented in the training data, it simply doesn't exist for those queries.
Models are trained on web crawls that prioritise: high-domain-authority pages with strong inbound link profiles; frequently cited sources across multiple independent publications; consistently accurate information that corroborates across many sources; clean, structured HTML that parsers can reliably extract; and author entities with verifiable real-world credentials. The single most powerful LLMO lever is getting your brand and expertise cited across multiple independent, high-authority sources — because that corroboration signals to both training pipelines and retrieval systems that your brand is a reliable source.
Brands cited in AI answers are 6.5× more likely to be cited through third-party sources than their own domain. This confirms what LLMO practitioners have observed: LLMs weight independent corroboration far more heavily than self-referential content. Your own website's content matters — but what others say about you matters more.
Every mention of your brand in a high-authority, independent publication strengthens your representation in AI training data. Target Wikipedia citations, academic references, government and regulatory mentions, major industry publications, and analyst reports. Quantity across diverse authoritative sources outperforms depth on your own domain.
Content that has historically been prioritised in training pipelines shares consistent traits: it is factually accurate and internally consistent; it contains verifiable claims with clear sources; it uses clean semantic HTML with proper heading hierarchy; it is cited by other high-authority sources; and it is written by identified, credentialed authors. Every piece of content you create should pass this filter.
LLMs struggle with ambiguous entities — brands, people, or concepts that share names or lack clear distinguishing attributes. Ensure your brand is unambiguously defined in Wikidata, Wikipedia (if applicable), Google's Knowledge Graph, and Schema.org markup on your own site. The clearer your entity definition, the more reliably models can reference you correctly.
Every description of your brand, product, or expertise across all platforms should be semantically consistent. Inconsistent descriptions (you're a "fintech platform" on your website but a "financial comparison tool" on LinkedIn and a "credit card aggregator" on Crunchbase) create ambiguity that reduces model confidence in representing you accurately.
GPT-5, observed in April 2026, conducts significantly more search fan-outs per query (10+ different sub-queries) and uses "site:" operators to get information directly from brand domains rather than third-party commentary. This suggests newer LLMs are moving toward a hybrid approach: corroborated third-party authority combined with direct brand source retrieval. Both channels must be strong.
LLMO measurement is still maturing, but a practical 2026 framework includes:
LLMO operates on a longer time horizon than GEO or SEO. Training data pipelines update on cycles of months. The brands investing in LLMO today — building corroborated authority, consistent entity signals, and training-data-grade content — are building a compounding asset that will become their dominant AI search advantage over the next 2–3 years.
Want to audit your LLMO position and build a long-term AI authority strategy?
Connect with Maheswari →Google's AI Mode has fundamentally changed organic search economics. Here's what gets cited, what still drives clicks, and how to position your brand for maximum AI Mode visibility in 2026.
Google AI Mode is a full conversational search experience built into Google Search, powered by Gemini. Launched in March 2025 and progressively rolled out globally through 2025–2026, it allows users to ask complex, multi-part questions and receive synthesised AI-generated answers — drawing on Google's search index through a retrieval-augmented generation (RAG) architecture.
Unlike AI Overviews, which appear as a box above traditional results, AI Mode replaces the traditional SERP entirely with a conversational interface. Users can follow up, ask for clarifications, and explore topics through multi-turn dialogue — all without visiting any external website. 93% of AI Mode sessions end without a single external click.
When a user asks a question in AI Mode, Google doesn't execute one search. It executes multiple sub-queries — "fan-outs" — to gather information from different angles. GPT-5 in April 2026 was observed making 10+ separate sub-query fan-outs for complex questions. Google's AI Mode operates similarly. This means your content doesn't need to rank for just one keyword — it needs to rank for the constellation of related sub-queries that AI systems generate around a user's intent.
A 2026 analysis by Dataslayer found that 92.36% of AI Overview citations come from domains ranking in Google's top 10 for that query. Domain authority is the single strongest predictor of AI citation — with high-authority domains earning roughly 3× more AI citations than lower-authority counterparts. The conclusion: Google's AI search is rooted in its existing ranking system. You cannot shortcut AI Mode visibility without first having strong traditional SEO fundamentals.
AI Mode supports multimodal responses — it can surface images, charts, diagrams, and video content alongside text. Content enriched with properly labelled images, structured data for visual elements, and video transcripts indexed by Google performs better in AI Mode than text-only pages, particularly for how-to and educational queries.
In May 2026, Google published a new documentation page titled "Optimising your website for generative AI features on Google Search." It is Google's most direct statement yet on AI search optimisation — and it contains several important clarifications that push back on vendor hype in the GEO/AEO space.
Google's guide confirms that its AI features are "rooted in our core Search ranking and quality systems." Foundational SEO — technical health, quality content, E-E-A-T, structured data, and link authority — is the foundation of AI search visibility. Google explicitly states: "From Google Search's perspective, optimising for generative AI search is optimising for the search experience, and thus still SEO."
Google's guide includes a new section on "agentic experiences" — AI agents that browse the web autonomously on behalf of users. Google frames this as "forward-looking rather than urgent" for most businesses — but for fintech brands where AI agents might compare rates, check loan eligibility, or research credit cards on a user's behalf, preparing structured, machine-readable product data is a non-trivial early-mover opportunity.
Core Web Vitals, crawlability, indexing, internal linking, quality content, and backlink authority are the foundation of AI Mode visibility. Brands that rank well in Google's traditional results will be cited in AI Mode. There is no shortcut around this foundation.
Experience, Expertise, Authoritativeness, and Trustworthiness are now even more critical for AI Mode than for traditional search. Author entity pages, credentials, bylines, expert review notes, and original research all signal E-E-A-T to Google's quality systems — and AI Mode inherits these quality signals directly.
The first sentence of any informational page should directly answer the primary question. AI Mode's retrieval system looks for "quick validation" — content that confirms the answer immediately before elaborating. Section headers should be questions or clear topic statements. Every section should be able to stand alone as a complete answer to its implied question.
Google has explicitly stated that chunking content into small pieces for AI readability is unnecessary. Write for human readers; Google's systems will handle the extraction.
Despite significant industry chatter, Google has confirmed these files receive no special treatment in Google Search. For non-Google AI platforms, the evidence of impact remains limited.
For financial services brands operating at scale, AI Mode creates both a threat and an asymmetric opportunity. The threat: informational content explaining financial concepts, eligibility criteria, or product features is highly vulnerable to AI Overview cannibalisation. The opportunity: comparison queries — "best credit card for travel rewards India", "which home loan has lowest processing fee" — require specific, current, verified data that AI struggles to synthesise accurately from generic sources.
Brands like BankBazaar that maintain live, structured, regularly updated product data — rates, fees, eligibility, features — are naturally positioned to be the authoritative cited source for commercial intent AI Mode queries, precisely because this data requires real-time accuracy that only dedicated fintech data teams can provide.
AI Mode is not the end of organic search — it's a restructuring. The organic click pie is smaller for informational content and larger for commercial content. Fintech brands that shift content investment accordingly, double down on data quality, and build AI citation authority will emerge from this transition with stronger, more defensible search positions than they had before AI Mode existed.
Want to audit your AI Mode readiness and build a citation-first SEO strategy?
Connect with Maheswari →100% of digital marketing practitioners surveyed in 2026 agree trust and credibility signals are growing more important — because AI systems now use brand authority as their primary filter for deciding which sources to surface. Here's the full playbook.
In traditional SEO, trust was one ranking factor among many — important, but balanced against relevance, freshness, and user signals. In AI search, trust has become the primary filter. AI systems don't just rank content — they decide which sources to include in answers at all. Brands they don't trust are simply absent.
A 2026 Goodfirms survey of 100+ digital marketing practitioners across 20+ countries found unanimous agreement: trust and credibility signals are growing more important as AI systems take on the work of deciding which sources to surface. This isn't a marginal shift — it's a structural change in how search value is allocated.
SE Ranking's 2026 analysis confirmed domain authority as the single strongest predictor of AI citation — with high-DA domains earning roughly 3× more AI citations than lower-DA counterparts. The mechanism is straightforward: AI systems are trained on web crawls that weight high-DA pages more heavily, and retrieval systems use authority signals to filter candidate sources. Building domain authority through quality backlinks and digital PR remains the most reliable AI citation lever available.
AI systems heavily weight independent corroboration. Brands cited across multiple high-authority, unaffiliated sources are far more likely to be included in AI answers than brands that only appear on their own domain. This is why digital PR — securing genuine earned media in authoritative publications — has become the single most important strategy for AI visibility, according to current industry consensus.
Google's E-E-A-T framework (Experience, Expertise, Authoritativeness, Trustworthiness) was designed for human quality raters but is now the primary framework AI systems use to evaluate sources. Every E-E-A-T signal you build — author credentials, bylines, expert review notes, original data, factual accuracy, transparent sourcing — directly strengthens your AI trust profile.
Inconsistent brand information across platforms creates entity ambiguity that reduces AI confidence. Consistent NAP (Name, Address, Phone), consistent brand descriptions, consistent author attribution, and consistent product/service descriptions across all platforms are fundamental trust hygiene for AI systems.
Digital PR — earning genuine editorial coverage in high-authority publications — has emerged as the most direct lever for AI search visibility. The logic: AI systems that rely on retrieval from the live web, and AI training pipelines that process web crawl data, both prioritise sources that are frequently cited by authoritative, independent publishers.
Newsworthy, data-backed pitches consistently outperform opinion-based or brand-promotion pitches for earning editorial coverage. Original research — proprietary data, platform-specific benchmarks, user behaviour studies — is the highest-value pitch asset because it provides journalists with exclusive, citable information they can't find elsewhere.
Claim and verify your Google Knowledge Panel. Build or improve your Wikidata entry with key entity attributes. Ensure consistent brand descriptions across LinkedIn, Crunchbase, industry directories, and social platforms. Establish author entity pages for your key content creators.
Execute monthly digital PR campaigns targeting Tier 1 publications. Publish original research quarterly — proprietary surveys, platform-specific data, industry benchmarks. Pursue podcast and speaking appearances in your industry vertical. Monitor unlinked brand mentions and convert to linked citations through proactive outreach.
Track AI citation rate across ChatGPT, Gemini, and Perplexity for your top 20 target queries. Identify topics where you are not being cited despite having strong content — these represent gaps in corroborated authority, not content gaps. Allocate digital PR resources to fill authority gaps by topic cluster, not just by domain.
| Metric | What It Measures | Tool / Method |
|---|---|---|
| AI citation rate | % of target queries where your brand is cited | Manual AI query monitoring; AI citation tools |
| Domain authority growth | Overall link authority trajectory | Ahrefs DR / Moz DA |
| Citation diversity | Number of distinct high-DA referring domains | Ahrefs / SEMrush referring domains report |
| Branded search volume | Awareness and brand recall proxy | Google Search Console branded query data |
| AI share of voice | Brand presence across AI answers relative to competitors | Manual competitive AI monitoring |
| Entity accuracy score | Consistency of how AI describes your brand | Multi-LLM brand description audit |
Brand authority compounds in a way that keyword rankings don't. Each piece of high-authority coverage strengthens your entity signals, which increases AI citation rate, which increases branded search and awareness, which drives more direct traffic and conversion — independent of any individual keyword ranking. This is why the brands winning in AI search in 2026 are the ones that started building genuine authority 2–3 years ago. The second-best time to start is today.
Want to build a compounding brand authority strategy for AI search visibility?
Connect with Maheswari →AI agents now browse the web on behalf of users — researching, comparing, and deciding without any human ever seeing your page. Google has included agentic guidance in its official documentation. Here's what Agentic SEO means and how to prepare.
Agentic SEO is the practice of optimising websites and content to be discoverable, interpretable, and trustworthy for AI agents — autonomous software systems that browse the web, gather information, compare options, and make decisions on behalf of human users.
In 2026, AI agents are no longer a speculative future. ChatGPT's operator-mode agents, Google's browser-integrated Gemini agents, Microsoft Copilot's agentic features, and numerous third-party AI agent platforms are actively browsing websites, reading product pages, comparing financial products, and synthesising purchase recommendations — all without any human user ever loading a page in a browser.
When an AI agent visits your product page, it doesn't appear in your standard analytics as a human session. It may arrive as a bot crawl, be blocked by JavaScript-dependent content, or access only machine-readable structured data. Your conversion funnel was built for humans. Agentic search requires you to build a parallel discovery layer for machines.
GPT-5, observed in April 2026, conducts 10+ search fan-outs per complex query and uses "site:" operators to pull information directly from brand domains. When a user asks ChatGPT to "find the best home loan in India with the lowest interest rate," GPT-5 may execute queries like: "site:bankbazaar.com home loan rates 2026", "HDFC home loan interest rate May 2026", "SBI home loan eligibility criteria", and several more — synthesising the answers into a single recommendation.
AI agents and web crawlers can reliably read: HTML text content with semantic structure; structured data (JSON-LD schema markup); server-side rendered page content; sitemap XML files; and robots.txt directives. They struggle with: client-side-only JavaScript-rendered content; dynamically loaded data that requires user interaction to trigger; content hidden behind authentication walls; and heavily Flash-dependent or non-standard web formats. A product page that requires JavaScript to render its interest rates is invisible to an AI agent doing product research on behalf of a user.
GPT-5's search behaviour confirms AI agents also look for authority signals specific to the query type. For "best nursing programs," the fan-outs included "NCLEX pass rates" and "CCNE accreditation." For "best SEO agency," they looked for Search Engine Land award winners. Agents are not just retrieving data — they are applying contextually appropriate authority heuristics to evaluate which sources and which data points to trust.
In May 2026, Google's new AI search documentation included a dedicated section on "agentic experiences" — the first time Google has officially addressed browser agents in its SEO documentation. Google frames agentic optimisation as "optional" and "forward-looking" rather than urgent for most businesses. But the inclusion signals that agentic search is now part of Google's strategic roadmap.
Google's guidance for sites where agentic access is relevant includes: ensuring key content is accessible to automated agents without JavaScript-only rendering; providing clear, structured, machine-readable data for products and services; and considering how agent-accessible structured data can support task completion for users who delegate research to AI assistants.
Google's documentation references "UCP" as a framework for agent-accessible web experiences. While the technical specification is still evolving, the principle is consistent with existing structured data best practices: provide enough machine-readable context that an agent can understand what you offer, at what terms, and whether it's a match for the user's need — without requiring a full browser session.
For fintech comparison platforms, agentic search is not a future consideration — it's a present reality with significant commercial implications. Consider this user journey: a user asks their AI assistant to "find the best credit card for international travel with no forex markup and a good lounge access benefit." The AI agent researches across multiple sources, synthesises a recommendation, and the user applies — never visiting a comparison website directly.
In this world, the comparison platform that wins is not the one with the best UX or the most SEO-optimised landing page. It's the one whose product data is most accurate, most machine-readable, most structured, and most efficiently retrievable by the agent conducting the research. Rate accuracy, data freshness, schema completeness, and server-side rendering of product specifications become the primary competitive levers.
Interest rates, fees, eligibility criteria, product features, and comparison data must be present in the initial HTML response — not loaded via client-side JavaScript. Use server-side rendering (SSR) or static site generation for product data pages. An AI agent that cannot read your interest rate table cannot include your product in its recommendation.
Use JSON-LD structured data to mark up all product data: FinancialProduct schema for credit cards, loans, and insurance products; Offer schema for current rates and fees; Review schema for user ratings; and FAQPage schema for eligibility and feature explanations. Machine-readable structured data is an agent's preferred input format.
Review your robots.txt for unintended blocks on legitimate AI crawlers. Consider allowing GPTBot, ClaudeBot, Googlebot (for Gemini agents), and Bingbot (for Copilot) access to product pages. Some brands are experimenting with blocking AI crawlers for training data purposes — but this also blocks agentic retrieval, which has a direct commercial cost.
Maintain HTML tables (not image-based or PDF-only) for current rates, fees, and feature comparisons. Update these on a schedule aligned with actual product changes. Freshness matters: an AI agent asked for "current home loan rates" will deprioritise stale data sources.
As GPT-5's behaviour demonstrates, agents apply category-specific authority heuristics. For fintech products, this includes: RBI registration and compliance signals, award recognitions from credible financial media, consumer trust ratings, and mentions in regulatory filings or official government resources. Identify the authority signals agents look for in your specific product category and proactively build them.
Agentic SEO is early-stage but accelerating. The brands investing in machine-readable product data, structured data completeness, and agentic authority signals today are building infrastructure that will become a decisive competitive advantage as AI agents handle an increasing share of user research and product discovery.
The transition from "search engine optimisation" to "agent optimisation" does not require abandoning everything learned in traditional SEO. Authority, accuracy, structure, and trust remain the foundational signals. What changes is who — or what — is consuming your content, and what they need from it. Build for the agent browsing your product page on a user's behalf, and you will build something better for human visitors too.
Agentic SEO is not optional for brands in high-consideration product categories. If an AI agent is researching credit cards, home loans, mutual funds, or insurance on behalf of a user, and your product data is not machine-readable, accurate, and structured — you are not in the consideration set. The conversion never happens. The traffic never appears. And you never know what you missed.
Want to audit your agentic readiness and structured data completeness?
Connect with Maheswari →Search no longer starts on a search engine for a huge share of users. Here's how discovery has fragmented across social platforms — and what optimising for it actually looks like.
For a large and growing segment of users, "search" no longer means opening a search engine. It means typing a question into TikTok's search bar to find a review, scrolling Instagram Explore for product comparisons, or searching a brand name plus "reddit" to find an unfiltered opinion. This isn't users abandoning search — it's search fragmenting across platforms that each serve a different kind of trust.
The common thread: users increasingly want content from people, not brands. A polished landing page competes poorly against a short video of someone actually using the product, or a Reddit thread where real customers argue about which option is genuinely better.
Social SEO is the practice of optimising content — video, captions, community posts — to be discoverable through each platform's own internal search and recommendation systems, rather than relying solely on traditional web search engines to drive discovery.
TikTok's search bar behaves increasingly like Google's — autocomplete suggestions, keyword-relevant results, and even AI-generated summaries for some queries. Optimising here means treating your video captions, on-screen text, and spoken audio as indexable content, since TikTok's algorithm reads all three.
Instagram's Explore and Search increasingly favour keyword-rich captions and alt text over hashtags alone. Product tags, location tags, and collaborative posts all feed into what content surfaces for a given query.
These aren't new, but their weight in Google's own results has grown sharply. Users trust community-sourced answers more than brand content for comparison and "is X worth it" queries — and Google's ranking systems have adjusted to reflect that trust.
Still the second-largest search engine globally. Long-form explainer and comparison videos with detailed, keyword-aligned titles and descriptions continue to rank both within YouTube and inside Google's video carousels.
Since Google's 2023–2024 core updates, Reddit and other community-forum threads have appeared dramatically more often in top organic results — particularly for comparison and recommendation queries. This reflects a broader signal: Google is rewarding content with visible, first-hand community consensus over single-author brand content for exactly the queries brands used to own.
You can't directly "optimise" a Reddit thread the way you optimise your own page. What you can do is participate authentically in relevant communities, encourage genuine customer discussion, and make sure your product information is accurate enough that when Reddit users do discuss it, the consensus works in your favour.
Traditional organic traffic reports miss most of this activity entirely, since it happens inside each platform. Track platform-native search impressions and saves (a stronger discovery signal than likes), branded search volume across platforms, and referral traffic from social apps to your site as a blended view of social search health.
Search hasn't shrunk — it's multiplied. The brands winning discovery in 2026 treat every platform where their audience asks questions as a search engine worth optimising for, not just a distribution channel for content built elsewhere.
Want to build a social search strategy alongside your core SEO?
Connect with Maheswari →Generating landing pages from structured data is one of the highest-leverage SEO plays available — and one of the fastest ways to trigger a quality update. Here's how to scale it responsibly.
Programmatic SEO means generating landing pages at scale from a structured data source and a page template — city + service pages, product comparison pages, IFSC/branch pages, or "best X for Y" combinations. Done well, it captures long-tail search demand that would take years to cover manually. Done poorly, it produces thousands of near-duplicate, low-value pages that Google's quality systems are specifically built to demote.
The failure pattern is consistent: a template with a swapped city or product name, but otherwise identical content, gets published across thousands of URLs. Google's helpful content systems are explicitly designed to detect this pattern — pages that exist primarily to rank rather than to serve a genuinely distinct user need.
Before publishing a programmatic page, ask: if a user landed here from search, would they get something they couldn't get from any other page in this set? If the honest answer is "not really," the page needs more unique data — or shouldn't exist.
A clear hub-and-spoke structure matters more at scale than on a handful of manually built pages. Build category hub pages that link out to relevant generated pages, and have generated pages link back to their hub and to genuinely related siblings — not an arbitrary "related pages" block. This gives crawlers a sane path through a large page set and helps distribute authority to pages that would otherwise sit several clicks deep.
On a large programmatic deployment, indexation ratio is the earliest warning signal available. Track indexed-vs-crawled-not-indexed in Search Console by page template, not just site-wide — a rising gap for one specific template is a direct signal that template's pages aren't clearing Google's quality bar, well before rankings or traffic show the damage.
| Signal | What It Indicates | Where to Check |
|---|---|---|
| Rising "crawled – not indexed" | Template quality below Google's bar | Search Console Pages report, filtered by URL pattern |
| Falling avg. time on page (by template) | Pages not matching user intent | GA4, segmented by template |
| Cannibalisation across near-duplicate pages | Insufficient differentiation between pages | Search Console query-to-page mapping |
| Sudden ranking drop across a template | Possible quality update impact on that page type | Rank tracking segmented by URL pattern |
Programmatic SEO isn't inherently risky — treating scale as a substitute for genuine usefulness is. The teams still winning with this approach in 2026 are the ones who scaled the data and the templates, but never scaled past the question of whether each page actually helps someone.
Want to scale programmatic SEO without risking a quality penalty?
Connect with Maheswari →Third-party cookies are effectively gone and privacy regulation keeps tightening. Here's the practical, non-theoretical playbook for building a first-party data strategy that actually holds up.
Third-party cookie deprecation, App Tracking Transparency, and an expanding patchwork of data protection regulation (GDPR, India's DPDP Act, and dozens of others) have permanently changed how marketers can identify and retarget users across the web. The brands still growing paid and personalised marketing performance made this shift years, not months, ago.
First-party data is information you collect from a user's behaviour on your own properties — purchase history, on-site activity, email engagement. Zero-party data is what a user tells you directly and intentionally — preferences stated in a quiz, a survey, or an account setting. Both are collected with consent and owned by you; both are becoming central to targeting and personalisation that used to rely on third-party tracking.
Zero-party data is often higher-quality signal for personalisation than inferred first-party behaviour — a user who tells you they're shopping for a home loan is a stronger signal than a user whose browsing history merely suggests it.
Collected data only has value once it's activated. Feed first-party data into platform "advantage/enhanced" audience-matching tools (which use hashed data without exposing raw PII), build lookalike and retention audiences from your own CRM data, and personalise on-site experience and email using the same profile — so a user's stated preferences carry across every channel they encounter your brand in, not just the one where they gave you the data.
Privacy-first marketing isn't a constraint to work around — it's a forcing function toward marketing that's more relevant and less wasteful. The brands treating consented first-party data as their core targeting asset, rather than a stopgap for lost third-party tracking, are the ones building a durable advantage.
Want to build a first-party data and privacy-first marketing strategy?
Connect with Maheswari →AI writing tools are now part of most content workflows. Here's how to use them without triggering the quality signals Google's systems are specifically built to catch.
Google has been consistent on this point: it doesn't penalise content for being AI-assisted, it penalises content that's unhelpful, regardless of how it was produced. The confusion comes from the fact that AI-generated content, published without meaningful editing or first-hand input, tends to score poorly on exactly the signals Google's helpful content systems were built to detect — generic phrasing, no original insight, no evidence of genuine experience.
The question was never "was AI used?" It's "does this add something a reader couldn't get from a dozen other pages on the same topic?" AI-assisted content can pass that test easily. Unedited AI output published at volume almost never does.
Use AI for what it's genuinely good at — first drafts, structural outlines, summarising research, generating variations to test — and route every piece through a human editor who adds the parts AI can't: a real example, a specific number from actual work, a judgment call informed by experience. The output should read like it was written by someone who could defend every claim in it under questioning, because that's effectively the bar Google's systems are approximating.
The strongest defence against being lumped in with generic AI content is concrete, specific, first-hand detail that a general-purpose model couldn't fabricate convincingly — a screenshot of your own dashboard, a number from a project you actually ran, a mistake you made and what you learned from it. This is also, not coincidentally, what makes content genuinely more useful to readers.
Adding a disclaimer that says "written with AI assistance" doesn't fix unhelpful content, and removing the disclaimer doesn't fix it either. Google's systems evaluate the content itself, not the production method disclosed or undisclosed.
AI tools are now a permanent part of the content workflow — resisting that is no longer a realistic strategy. The differentiator isn't whether you use AI, it's whether a real person with real expertise stood behind every piece before it published.
Want to audit your content workflow for E-E-A-T resilience?
Connect with Maheswari →The marketers scaling fastest in 2026 aren't working more hours — they've built AI-agent pipelines that handle the reporting, tracking, and first-draft work automatically. Here's what to automate first.
Weekly rank tracking, monthly performance decks, keyword clustering, and first-draft content briefs used to consume a huge share of a marketing team's time — not because the work was hard, but because it was repetitive and manual. Workflow automation platforms combined with AI models have turned most of this into infrastructure that runs on a schedule instead of a to-do list.
A workflow automation tool (like n8n) acts as the connective layer between data sources — Search Console, GA4, SEMrush or Ahrefs APIs, Google Sheets — and an AI model that summarises or triages what comes through. The pattern that works well: scheduled trigger → pull data from source APIs → AI model summarises or flags anomalies → structured output pushed to Slack, email, or a shared sheet. None of the individual pieces are exotic; the value is in wiring them together so the whole pipeline runs unattended.
Automation handles the parts of marketing that are repetitive and rule-based well. It handles poorly the parts that require judgment — deciding which anomaly actually matters, which content gap is worth pursuing, which brand voice choice fits a specific audience. The teams getting the most value from automation are the ones using the time it frees up for exactly that kind of judgment work, not the ones trying to automate the judgment itself.
The goal of marketing automation isn't fewer marketers — it's marketers spending their time on the 20% of the work that actually requires a human decision, instead of the 80% that never did.
Want to build an AI-agent automation pipeline for your marketing team?
Connect with Maheswari →