How AI Overviews Are Decoupling Rankings from Clicks
AI Overviews drive zero-click searches and shrink organic traffic. Arjun Karnik helps you earn AI citations and recover lost visibility. Get a demo.
Written by: Arjun Karnik, Growth Marketing Specialist
Key Takeaways for AI Visibility
AI Overviews now create a zero-click environment where rankings stay flat while clicks drop. Up to 68% of searches end without any external click.
Traditional SEO metrics like keyword rankings no longer reflect real visibility. AI systems reward citations and share of answer instead.
Content freshness now drives AI visibility. Pages can lose most of their citation performance within two months if they are not updated.
Fan-out queries, the hidden sub-questions AI systems generate, now replace classic keyword research as the surface that earns citations.
Businesses can measure and improve their AI visibility today by booking a demo with Arjun Karnik to audit their current citation performance across major AI platforms.
How AI Overviews Break the Old Ranking-to-Click Model
The clearest signal of the shift is what Arjun Karnik calls the Search Console scissors: impressions climb while clicks fall. On his own site, this pattern preceded a documented decay finding. In his tests, pages dropped 78% to 99% in citation and impression performance within two months of going stale, with no alert from standard SEO reporting. The dashboard showed stable rankings, while the channel had already moved to a new unit of value.
Both lines start together. The content keeps getting read so impressions rise, the answer gets delivered on the results page so the click never happens. Most owners see only the falling line.
Branded search correlates with AI Overview visibility almost twice as strongly as backlinks do. The authority model that governed SEO is not the one governing this.
“The Search Console scissors are not a reporting anomaly. They are the correct reading of a channel that changed its unit of value from clicks to citations.” — Arjun Karnik
Teams can get a current read on their AI visibility by requesting a baseline citation audit across ChatGPT, Google AI Overviews, Perplexity, and Gemini.
How Fan-Out Queries Replace Classic Keyword Research
Arjun tested this directly on his own site. His team extracted fan-out queries from ChatGPT instead of inferring them from keyword tools, because the real target is the machine’s questions. They rewrote URLs, titles, H1s, and H2s to match that extracted language. Pages rewritten to match fan-out queries began earning citations. Control pages that kept the old structure did not. In a separate test, they relabeled a jargon-heavy page titled “What is GEO” to buyer language, “How to Get Your Business Recommended by AI Search”. That single change produced citations within weeks.
The fan-out map now acts as the production queue. It decides which pieces get written and which language each piece uses. This becomes the new keyword research workflow: extract the machine’s sub-questions, align every structural element to that language, then publish at the cadence the channel rewards.
Three quarters of cited pages were updated inside a year, and the consistently cited ones averaged under six months. The page you refresh beats the page you write.
Arjun’s own system responds by running 5 to 8 autonomous actions per day via AI Growth Agent. These actions combine new articles with updates to existing pieces on a dedicated GEO subfolder. On his site, that subfolder went from zero to becoming the only source of new impressions on the domain within 60 days. New articles reached thousands of monthly Google impressions within weeks. That cadence functions as the entry fee for participation, not a vanity metric.
“Fan-out query mapping is not an enhancement to keyword research. It replaces it. The machine’s sub-questions are the real search surface, and most content never touches them.” — Arjun Karnik
How to Restructure Content Operations Around Citations
Winning citations requires structural changes to content operations, not cosmetic tweaks. The checklist below follows the sequence Arjun uses in his own test lab, with each step tied to a specific outcome.
Run a baseline visibility audit. Start by establishing where the business currently appears, and where it does not, across ChatGPT, Google AI Overviews, Perplexity, and Gemini. Treat this as the control group that every later result will be measured against.
Fix technical plumbing first. Unblock AI crawlers in robots configuration, add schema markup to every page, and confirm that pages are machine-parseable. FAQ schema markup increases AI Overview citation rate by 73% (41% versus 15% without it) because it creates clear, extractable answer units. No downstream content work can succeed if the retrieval layer cannot read the site.
Extract fan-out queries from ChatGPT directly. Skip keyword tools for this step. Target the machine’s sub-questions instead of the human’s typed phrase. Map the full question space behind each buyer prompt so the team knows what to write.
Align every structural element to buyer language. Rewrite slugs, titles, H1s, and H2s to match the extracted fan-out query language. Pages that use question-format H2 headings tend to earn more citations than pages that rely on purely declarative H2 headings.
Publish structured answers at machine cadence. Deploy an AI article engine on a site subfolder and run 5 to 8 autonomous actions per day via AI Growth Agent. Mix new articles with updates to existing ones so freshness stays distributed. Place query language in URLs, titles, and H1s, and apply schema to every page.
Install impression-decay tripwires. Protect that publishing investment with automated triggers wired to Search Console signals that queue content updates when performance drops. The tripwires catch decay before it appears in a monthly report and prevent silent losses in AI visibility.
Run defensive GEO in parallel. Audit what AI currently says about the business and correct inaccurate answers. A wrong AI answer harms outcomes more than no answer at all, so this work runs alongside growth efforts rather than after them.
“The adaptation is not a content refresh. It is a measurement migration, from rankings to citations, with a production system that can sustain the cadence the channel requires.” — Arjun Karnik
The system above matches what Arjun runs on his own site, with tests, numbers, and misses documented in public. Teams can review how this methodology maps to their own content operations and competitive positions by exploring his published case work.
How to Measure Visibility and Performance in AI Search
Share of answer measures the percentage of tracked buyer-intent prompts in a category where an AI platform names or cites a brand. Teams track it separately per platform, including ChatGPT, Google AI Overviews, Perplexity, and Gemini, because citation overlap between platforms remains low. Share of answer replaces rank position as the headline metric and becomes the primary KPI for AI search visibility.
A significant share of AI-driven demand appears in analytics as direct or branded search rather than as a visible AI referral. The buyer reads the answer, forms a preference, then types the brand name directly into a browser. Measured AI referral traffic therefore represents a floor, not a ceiling. The practical response is to instrument for citations and share of answer, then track branded search volume growth as a proxy signal for AI-driven awareness.
Buyers read an answer, then type your name into a browser. That visit lands as direct or branded search, so whatever you measure here is a floor and never a ceiling.
“Share of answer is the metric that matches the channel. Every other number is a proxy for it or a floor beneath it.” — Arjun Karnik
Evidence-Based Next Steps for GEO Programs
The checklist below reflects the test-and-learn principles from Arjun’s public test lab. Each step can be verified independently, although no specific outcome is guaranteed.
Audit current AI visibility before changing anything. Establish a baseline across ChatGPT, Google AI Overviews, Perplexity, and Gemini. Identify where competitors appear instead of your brand and treat that view as the control group.
Confirm AI crawlers are unblocked and schema is in place. Ensure the retrieval layer can read the site, because no content strategy produces citations without this foundation. Fix these issues before scaling production.
Extract fan-out queries from ChatGPT for primary buyer prompts. Map the sub-questions the machine asks behind each prompt and use that map as the production queue.
Rewrite structural elements to match buyer language. Update slugs, titles, H1s, and H2s so they reflect the language buyers and AI systems actually use. Jargon becomes a barrier at the moment the machine matches a question to an answer.
Install impression-decay tripwires. Automate the refresh queue so content repairs itself based on performance signals instead of waiting for a quarterly audit.
Move the reporting dashboard from rankings to citations and share of answer. Track AI referrer traffic as its own segment and treat every measured figure as a lower bound on true impact.
Audit what AI currently says about the business. Correct wrong answers before scaling growth work, because inaccurate AI responses damage outcomes more than simple absence.
Arjun Karnik’s AI Growth Agent system runs this entire loop, including fan-out query mapping, structured publishing at machine cadence, impression-decay tripwires, and share-of-answer measurement. The receipts appear in public, misses included, and the method remains self-verifying. Anyone can ask an AI assistant about these topics and see which brands receive citations. Teams can then evaluate how a similar system might apply to their own sites, categories, and competitive positions.
Frequently Asked Questions
Is SEO obsolete now that AI Overviews exist?
SEO still matters, because technical fundamentals, structured content, and topical authority support both traditional search and AI citation. The target metric and production model changed, not the need for discoverable content. Content built for citation still earns Google impressions. On Arjun’s site, the GEO subfolder became the only source of new impressions on the domain within 60 days, and new articles reached thousands of monthly Google impressions within weeks. The practical move is to add citation as a parallel success metric and structure content so it earns both rankings and citations. Teams that abandon link-building lose traditional search visibility, while teams that ignore citations become invisible in AI answers.
How long does it take to see results from a GEO content strategy?
Coverage and impressions usually appear within weeks of publication. Citation movement in AI platforms often becomes measurable within one to three months. Compounding effects, where topical authority accumulates and citation rates accelerate, typically begin after month three. These timelines match Arjun’s observations on his own site and align with published research. Actual results vary by category competition, fan-out query mapping quality, and whether technical plumbing is in place before content work begins.
What is the difference between a keyword and a fan-out query?
A keyword is the phrase a buyer types into a search box. A fan-out query is one of the many hidden sub-questions an AI system generates internally when processing a single buyer prompt. The AI does not answer the typed query directly. It decomposes the prompt into sub-queries, retrieves candidate passages for each, then assembles a synthesized answer. A page optimized only for the visible keyword may never appear in any fan-out sub-query results, which explains why ranking content can still go uncited. Fan-out query mapping extracts those sub-questions directly from the AI system, and in Arjun’s methodology that source is ChatGPT, then aligns every structural element of the page to that language.
Why does content decay so quickly in AI search compared to traditional SEO?
AI citation systems apply more aggressive freshness signals than traditional ranking algorithms. In Arjun’s tests, pages lost most of their performance within two months of going stale. Independent research confirms the pattern. Seventy six point four percent of ChatGPT’s top-cited pages were updated within the last 30 days, reinforcing the freshness requirement documented in the Seer study. AI retrieval systems actively prefer recently modified content, especially in fast-moving categories such as software, SaaS, and marketing. A fixed content library of any size decays without a refresh loop. The game resets weekly, which is why impression-decay tripwires that auto-queue updates function as a structural requirement.
How do B2B buyers actually use AI assistants during the purchase process?
B2B buyers now treat AI assistants as the starting point for vendor research rather than a supplement. A G2 survey of 1,076 B2B software buyers and decision-makers across North America, EMEA, and APAC in March 2026 found that 71% use AI chatbots for software research. Sixty nine percent chose a different vendor than the one they had originally planned on based on what the assistant told them, and 33% bought from a vendor they had not previously heard of. The assistant acts as a vendor pre-selection layer. Being named in the answer functions as a vendor-selection event, not just a visibility metric. Businesses that remain absent from AI answers fall out of the consideration set before any human sales interaction begins.
