Written by: Arjun Karnik, Growth Marketing Specialist

Why Impressions Are Rising While Clicks Fall

  • AI Overviews and zero-click features now absorb many searches. Pew Research shows clicks drop from 15% to 8% when AI summaries appear.
  • 71% of B2B buyers use AI chatbots for research, and 69% switch vendors based on AI recommendations, so citations now matter more than rankings.
  • Content freshness drives visibility. Seventy-five percent of AI-cited pages were updated within the last year, and pages can lose 78–99% of visibility in two months without updates.
  • Targeting fan-out queries, the 8–10 hidden sub-queries per prompt, and matching buyer language significantly increases AI citation rates.
  • See the fan-out query mapping methodology in action and learn how Arjun Karnik’s test-lab approach shifts measurement from clicks to citations.

Why “Impressions Up, Clicks Down” Happens

Impressions up, clicks down means your content ranks, AI systems use it to construct answers, and the user never visits your site. The impression registers. The click does not.

Line chart showing the scissors pattern over twelve months, with an impressions line rising while a clicks line falls away from it. Illustrative shape of the pattern, not data from a specific account.
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.

AI Overviews now appear in approximately 48% of Google search results as of 2026. Seer Interactive’s April 2026 analysis of 53 brands, 5.47 million queries, and 2.43 billion organic impressions found continued CTR declines for organic searches with AI Overviews present.

Bar chart comparing click-through rate on a traditional search result, 15 percent with no AI summary shown and 8 percent when an AI summary is shown. Source: Pew Research Center, July 2025, 900 US adults across 68,879 Google searches.
The click roughly halves when an AI summary appears above the result. Pew also found only 1 percent of users clicked a link inside the summary itself.

A single buyer prompt triggers a cluster of lookups across the web. Ekamoira research analyzed 72,000+ AI-generated queries and found that a single prompt in ChatGPT or Gemini routinely triggers 8–10 parallel sub-queries before an answer is returned. Optimizing for the visible prompt while ignoring the fan-out targets the wrong surface. Content can rank and still go uncited.

Surfer SEO’s study found pages ranking for fan-out sub-queries are 161% more likely to be cited in Google AI Overviews; a separate Ekamoira study examined 173,902 URLs on related AI-citation factors. Surfer SEO’s study also found that 67.82% of pages cited in AI Overviews do not rank in the traditional top 10 search results for the query.

Freshness now acts as a primary ranking factor for AI systems. Seer Interactive analyzed 47,097 AI citations across 7,683 pages in ChatGPT, Gemini, and Perplexity between March and June 2026 and found that 75% of cited pages had been updated within the last year, with pages cited consistently across all four months averaging under six months since their last update. In Arjun’s own decay tracking on his site, pages can drop 78% to 99% in two months without updates, long before a monthly report flags the loss.

Bar chart showing 75 percent of pages cited by AI assistants were updated within the last year and 25 percent were older. Source: Seer Interactive, July 2026, 7,683 pages and 47,097 citations across ChatGPT, Gemini and Perplexity.
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.

76.4% of pages cited by ChatGPT were updated within the prior 30 days. Ahrefs analysis of 17 million citations found that AI-cited content is 25.7% fresher than traditional organic Google results.

The zero-click path also hides real impact. A meaningful share of AI-driven demand lands in analytics as direct or branded search rather than anything traceable to the answer that caused it. Whatever you measure is a floor, not a ceiling.

Bar chart showing 2.5 percent of downstream brand visits after an AI mention carry a trackable referral parameter while 97.5 percent arrive untraceable. Source: Profound, analysis of more than 2 million AI conversations, January to June 2026.
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.

From Impressions to Influence: A 7-Step AI Visibility Plan

The real goal is earning mentions, citations, and recommendations in the answers buyers read before they ever visit a site. The 7-step checklist below comes from Arjun’s own test lab on his site, using AI Growth Agent.

  1. Run a visibility baseline first. Audit what ChatGPT, Gemini, Perplexity, and Google AI Overviews currently say about your business before publishing a single new page. A wrong AI answer hurts more than no answer.
  2. Fix technical plumbing. Unblock AI crawlers, add schema markup to every page, and make pages machine-parseable. Nothing downstream works if the retrieval layer cannot read the site.
  3. Extract fan-out queries directly from ChatGPT. Skip keyword tools for this step. The target is the machine’s questions, not the human’s visible prompt. As noted earlier, fan-out queries create multiple retrieval opportunities per prompt, and pages aligned to these sub-queries achieve a 51% AI citation rate compared with 20% for main-query-only content.
  4. Align to buyer language, not practitioner jargon. In a test on Arjun’s site, a page titled “What is GEO” was relabelled “How to Get Your Business Recommended by AI Search,” with the slug, title, H1, and H2s all realigned to buyer questions. Citations followed within weeks of that specific change.
  5. Rewrite URLs, titles, H1s, and H2s to match the fan-out map. In Arjun’s tests, pages rewritten to match extracted ChatGPT fan-out queries earned citations while control pages did not.
  6. Publish structured pages at machine cadence. Query language belongs in URLs, titles, and H1s because AI retrieval systems check those fields first. Schema markup belongs on every priority element so those signals become machine-readable. Answer-first formatting acts as a structural requirement that determines whether your content gets parsed at all.
  7. Install impression-decay tripwires. Set automated triggers against Search Console signals that queue an update when performance drops. Given the decay rates documented earlier, these tripwires catch performance issues before positions disappear entirely.

The measurement target shifts at step one and stays shifted. The buyer verbs that matter now are mentioned, recommended, and cited, not ranked.

Three Levers That Turn AI Visibility Into Pipeline

Buyer-language alignment, structured publishing, and freshness loops convert AI visibility into downstream traffic and pipeline. Each lever is independently testable on your own site.

Buyer-language alignment. Jargon blocks relevance at the moment the machine matches a question to an answer. Pages labelled in the words buyers use, not the words practitioners use, earn more citations. The slug, title, H1, and every H2 should reflect the question the buyer typed into the assistant, not the internal category term.

Structured publishing. The Princeton and Georgia Tech GEO paper found content with original statistics receives 30–40% higher AI visibility. Structure acts as a retrieval requirement. Prose that is not structured for extraction, not carrying schema, and not aligned to fan-out query language will lose to a worse-written page that satisfies all three conditions.

Freshness loops. On Arjun’s site, the GEO subfolder went from zero to the only source of new impressions on the domain in 60 days. AI Growth Agent ran a steady cadence of new articles and updates. New articles reached thousands of monthly Google impressions within weeks. Seer Interactive testing found that refreshing outdated content led to a 300% increase in AI traffic for one client and a 54% increase in GPT-User bot hits for another.

Approximately 50% of sources cited for a given prompt will change within 13 weeks. A fixed library of any size decays in place. The refresh loop becomes the entry fee, not optional maintenance.

Attribution will still understate the result. The zero-click path runs answer, then brand search, then visit, which shows up in analytics as direct or branded traffic. Measure citations and share of answer as the primary signal, and treat clicks as a floor.

Explore how freshness loops and fan-out mapping would work on your site and see AI Growth Agent’s system applied to your domain.

Metrics That Replace Clicks in an AI-First World

Citation and share-of-answer measurement across ChatGPT, Google AI Overviews, Perplexity, and Gemini now replace rank position as the headline metric. The table below compares the SEO and GEO measurement frameworks for a $1M–$20M B2B business.

Dimension SEO GEO
Primary success metric Rankings and organic clicks Citations, mentions, and share of answer
Query model The keyword the buyer typed Clusters of hidden fan-out sub-queries triggered by one prompt
Authority signal Backlinks and domain authority Topical coverage; brand web mentions correlate with AI Overview visibility at a Spearman coefficient of 0.664 vs 0.218 for backlinks
Freshness requirement Periodic; accumulated authority sustains rankings Continuous; median citation half-life is 4.5 weeks across AI platforms
Click-through reality Position 1 historically drove 28%+ CTR Being cited in an AI Overview delivers 120% more organic clicks per impression than not being cited, yet still produces 38% fewer clicks than queries with no AI Overview present
Where wins compound Domain authority accumulates over years Citation concentrates around 3–9 sources per query across most major AI engines, though ChatGPT averages about 15.

Share of answer is measured by running a fixed set of 20–50 target prompts weekly against ChatGPT, Perplexity, Google AI Overviews, and Gemini. You log citations per engine and calculate the percentage of answers that name the brand or domain. AI referrers such as chatgpt.com are tracked as a distinct traffic class in analytics because ChatGPT-referred visitors spend more time on site and convert at higher rates than typical Google search traffic. They behave like referrals, not like cold search traffic.

AI visibility screen filtered to Google AI Overviews, showing a mention rate trend chart climbing over time and crossing above a dashed competitor benchmark line, with range controls and tabs for overview, wins, position trends and top URLs.
Mention rate over time against a competitor benchmark. This is the number that replaces rank position. The figures shown are a product view, not a client result.

AI Growth Agent clients average more than 12,000 additional AI citations and mentions, over 100,000 additional bot visits, and a 20% or greater lift in impressions across the first twelve weeks. Those are AI Growth Agent’s published figures, not Arjun’s, and are cited as such.

Next Steps to Test This on Your Own Site

The following checklist reflects the test-and-learn sequence Arjun runs on his own site. Every item is independently verifiable on your own property.

  1. Pull your Search Console scissors chart today. Compare impressions and clicks over the last 12 months on informational queries. If impressions rise while clicks stay flat or fall, the great decoupling is already active on your domain.
  2. Run a visibility audit across all four surfaces. Ask ChatGPT, Gemini, Perplexity, and Google AI Overviews the 10 questions your best buyers ask before they buy. Record who gets cited. If your name does not appear, that becomes the baseline to beat.
  3. Fix technical plumbing before anything else. Confirm AI crawlers are unblocked. Add schema markup. Make every priority page machine-parseable. This silent blocker can make every downstream investment irrelevant.
  4. Extract fan-out queries from ChatGPT for your top 5 topics. Skip keyword tools for this step. Type your buyer’s most common prompt into ChatGPT and map the sub-questions it generates. That map becomes your production queue.
  5. Rewrite one page to buyer language and fan-out alignment. Change the slug, title, H1, and H2s to match the extracted questions. Hold a control page back. Measure citation difference over 30 days. This is the test Arjun ran on his site that produced citations on rewritten pages while controls stayed uncited.
  6. Set a freshness tripwire on your highest-impression pages. Define a drop threshold in Search Console. When a page crosses it, queue an update immediately. In Arjun’s tests, waiting for a monthly audit means the position is already gone.
  7. Shift your primary reporting metric to share of answer. Track citations per engine weekly. Track AI referrers as a separate segment in analytics. Stop grading this channel on clicks alone. The buyer journey now runs answer, then brand search, then visit, and the click in the middle is often the step the buyer skips.

The window for outsized gains is open now. A small number of brands capture the majority of AI responses per category. Early citations become tomorrow’s settled answers, and settled answers tend to stick.

Review Arjun’s full test-lab methodology on your domain and see fan-out query mapping, freshness loops, and citation measurement applied to your site.

Frequently Asked Questions

What does “impressions up, clicks down” actually mean in Google Search Console?

“Impressions up, clicks down” means your pages appear in search results more often, but fewer users click through to your site. Google Search Console registers an impression every time your URL loads in a results page, whether or not the user scrolls to see it or clicks it. A click only registers when the user selects your link and leaves the results page.

The divergence between these two numbers comes primarily from AI Overviews and other zero-click features that answer the query directly on the results page. The user reads the answer, gets what they need, and never visits your site. Your content did the work. The click did not follow, because the journey now runs from AI answer to brand search to direct visit rather than from query to article click to conversion.

Is the impressions up, clicks down pattern a sign that my SEO is failing?

The pattern signals a channel shift, not automatic failure. If your impressions rise on informational queries while clicks fall, AI systems are likely consuming your content to construct answers. That counts as a visibility event, not a ranking failure.

The failure mode appears when you keep measuring the channel by clicks alone while the buyer journey adds a step the click metric cannot see. The correct diagnostic is to check whether your brand is being cited in those AI answers. If you are cited, the content is working and the measurement is wrong. If you are not cited, the content is working for a competitor’s answer and the strategy needs to change.

The scissors chart in Search Console shows the visible half of the problem. The invisible half is the share of AI answers where your name does and does not appear.

What is fan-out query mapping and why does it matter for AI search visibility?

Fan-out query mapping identifies the hidden sub-queries that an AI assistant generates underneath a single buyer prompt before constructing its answer. When a buyer types a question into ChatGPT or Perplexity, the system does not perform a single lookup. It generates multiple parallel retrieval queries and assembles the answer from what comes back across all of them.

If your content targets only the visible prompt and ignores the sub-queries underneath it, the retrieval layer may never surface your pages, even if you rank well in traditional search for the head term. Fan-out mapping means extracting those sub-questions directly from the AI system, then aligning your URLs, titles, H1s, and H2s to match that language.

In a test on Arjun’s site, pages rewritten to match extracted fan-out queries earned citations while control pages did not. The fan-out map also becomes the production queue. It tells you what to write next and which language to use.

How do I measure success if clicks are no longer a reliable metric?

Success now centers on citations and share of answer across ChatGPT, Google AI Overviews, Perplexity, and Gemini. Share of answer is calculated by running a fixed set of target prompts, typically 20 to 50, on a weekly or biweekly basis, logging which engine cites your domain, and calculating the percentage of answers that include your brand or URL.

Alongside that, AI referrers such as chatgpt.com are tracked as a distinct segment in your analytics platform. Traffic arriving from AI assistants converts differently from cold search traffic and should not be blended with it. Google Search Console remains useful for tracking impressions and decay curves, but the headline metric no longer centers on rank position or click volume.

One honest caveat applies. Buyers frequently copy an answer from an AI assistant and paste a brand name directly into a browser, which shows up in analytics as direct traffic and never gets attributed to the AI answer that caused the visit. Whatever citation and referral numbers you measure are a floor, not a ceiling.

How quickly does content decay in AI search, and how often should I update pages?

Decay moves faster than most content calendars anticipate. In Arjun’s tests on his site, pages can drop 78% to 99% in two months without updates. Independent research points the same direction. Pages updated within the last 30 days earn substantially more AI citations than older content, and the median citation half-life across major AI platforms is approximately 4.5 weeks.

The practical implication is that a fixed content library of any size decays in place. The game resets weekly, so volume and cadence stop being vanity metrics and become the entry fee. High-value commercial pages and comparison content require the most frequent updates, on cycles of 30 to 60 days.

Evergreen definitional content can sustain longer cycles, but no priority page should go more than six months without a substantive update if AI citation is a goal. The most efficient approach is to set impression-decay tripwires in Search Console that automatically queue an update when a page’s performance drops past a defined threshold, rather than waiting for a quarterly audit to surface the problem after the position is already lost.