Impressions Up, Clicks Down: Fix the Scissors Pattern

Written by: Arjun Karnik, Growth Marketing Specialist

Key takeaways from the impressions-up-clicks-down pattern

• The “impressions up, clicks down” pattern signals AI Overviews and chatbots consuming your content without sending traffic, not weak content.
• Four root causes drive this scissors pattern: AI Overviews resolving queries on SERPs, hidden content decay, intent misalignment, and technical blocks that stop AI crawlers.
• Mapping hidden fan-out queries behind buyer prompts and matching buyer language can increase AI citation rates by up to 2.2x.
• Technical fixes such as unblocking AI crawlers, adding schema markup, and using answer-first formatting must come before any content changes.
• Get a custom diagnostic built around your Search Console data

How to read “impressions up clicks down” in Search Console

Four numbers define this pattern: impressions, clicks, click-through rate (CTR), and average position. When impressions rise while clicks fall, CTR collapses even if position holds. That combination of stable or improving position, rising impressions, and falling CTR is the diagnostic signature of AI consumption without referral.

The filter table below shows how to isolate the pattern in Google Search Console’s Performance report.

View	Filter	What to look for	Signal
Queries	Date range: current 90 days vs. prior 90 days	Impressions up, clicks flat or down, CTR falling	AI Overview likely present on these queries
Pages	Date range: current 90 days vs. prior 90 days	Pages with rising impressions and falling clicks	Content being consumed without referral
Search type	Web only	Informational queries with CTR below 1%	High AI Overview exposure; Seer Interactive found organic CTR on AI Overview queries fell from 1.76% to 0.61%
Queries	Filter by query type: “what,” “how,” “why”	Informational head terms with high impression volume	Most exposed to zero-click behavior; informational queries show a 71% zero-click rate

The scissors chart, with impressions climbing left and clicks falling right, reflects channel mechanics rather than content quality. SparkToro’s analysis of Similarweb clickstream data found that 68.01% of US Google searches ended without a click in January through April 2026, up from 60.45% in 2024. The buyer read the answer where they asked it. The click you are missing is not a lost ranking, it is a step the buyer skipped. Understanding this pattern is the first step; the next step is identifying what drives it on your site.

Four root causes behind rising impressions and falling clicks

Four root causes produce the scissors pattern, and each one is diagnosable from Search Console data.

AI Overviews and rich snippets resolve queries on the SERP. AI Overviews now appear in approximately 48% of Google search results as of 2026. The Pew Research Center tracked 68,879 Google searches across 900 US adults in March 2025 and found users clicked a traditional result in 8% of visits when an AI summary appeared, versus 15% without one. That shift removes roughly half the clicks on queries where AI Overviews trigger. SparkToro found AI Overviews reduce click-through rates by nearly 60% when present.

Content decay stays invisible until position finally drops. In Arjun Karnik’s own tests on his site, pages dropped 78% to 99% in two months without updates. Standard monthly rank reports rarely catch that decay before it compounds. 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 consistently cited pages averaging under six months since their last update.

Intent misalignment answers the wrong version of the question. A single buyer prompt triggers multiple hidden retrieval queries underneath it. An AirOps analysis of 548,534 retrieved pages across 15,000 original prompts found that ChatGPT generates two or more follow-up searches on 89.6% of queries, with 95% of those fan-out queries showing zero monthly search volume in conventional keyword tools. Content tuned only to the visible keyword misses the retrieval surface where AI systems actually work.

Weak technical plumbing blocks the retrieval layer. When AI crawlers are blocked in robots.txt, when pages carry no schema, and when content appears as unstructured prose, the machine cannot reliably read the page. The quality of the writing does not matter until the retrieval layer can access and parse it. This silent blocker often makes every other fix irrelevant.

See how these root causes apply to your content library

Mapping hidden fan-out queries behind buyer prompts

Fan-out query mapping starts with the machine’s questions, not the human’s. Keyword tools report search volume for queries people type. They do not report the sub-queries an AI system runs internally before assembling an answer. Those sub-queries define the actual retrieval surface that decides which pages get cited.

Capturing these hidden sub-queries requires a systematic approach that forces the AI to reveal its internal retrieval logic. The workflow Arjun uses on his own site runs as follows.

1. Open ChatGPT and enter the buyer prompt you want to rank for, for example, “how do I get my business recommended by AI search.”
2. Ask ChatGPT to list every sub-question it would need to answer to respond fully to that prompt, then capture the full list.
3. Repeat with three to five variant phrasings of the same prompt. Between any two runs, as few as 27% of fan-out strings remain identical depending on the AI platform, so multiple runs are required to identify reliable themes.
4. Cluster the sub-questions by topic so each cluster becomes a content target.
5. Audit existing URLs against the cluster list. For each gap, create a new page. For each existing page that partially covers a cluster, rewrite the slug, title, H1, and H2s to match the buyer-language version of the question.

The buyer-language step changes performance, not just cosmetics. In a documented test on Arjun’s own 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. Jargon blocks relevance at the exact moment the machine matches a question to an answer.

Pages with 50% or greater title-to-query overlap achieved a 20.1% citation rate in ChatGPT versus 9.3% for pages with less than 10% overlap, a 2.2x lift from alignment alone. Studies show that pages covering multiple sub-queries can achieve higher AI citation probability than traditional SEO pages that target only the head term.

Freshness and measurement loops that slow content decay

Freshness functions as the entry fee for AI visibility, not a hygiene chore. In Arjun’s tests on his own site, pages lost 78% to 99% of their performance in two months without maintenance. That decay stays hidden in a monthly rank report because position can hold while citation share collapses.

The freshness loop Arjun runs via AI Growth Agent uses impression-decay tripwires. These are automated triggers wired to Search Console signals that queue a content update when a page’s performance drops past a set threshold. The threshold is calibrated against the decay curves measured in those tests, where pages lost most of their performance in under 60 days. The update queue fires without anyone auditing a spreadsheet.

On Arjun’s own site, the GEO subfolder went from zero to the only source of new impressions on the entire domain in 60 days. New articles reached thousands of monthly Google impressions within weeks. In a controlled test, pages rewritten to match extracted fan-out queries earned citations while control pages did not.

Citation measurement runs in parallel across four surfaces: ChatGPT, Google AI Overviews, Perplexity, and Gemini. The measurement framework tracks three tiers of outcome per prompt.

• Mention: the brand name appears in the response with no linked source.
• Citation: a URL from the site is named as a source.
• Recommendation: the assistant actively suggests the brand as the answer.

Share of answer, the percentage of tracked prompts where the brand appears, replaces rank position as the headline metric. Strong B2B SaaS programs can achieve significant share of answer on targeted prompts, while weaker programs often see low results even when they hold high organic rankings.

AI referrer tracking in analytics, where chatgpt.com and equivalents appear as a distinct traffic class, closes more of the attribution loop. One honest caveat still applies. Buyers often copy an answer and type a brand name directly into the browser, which lands as direct traffic. Whatever you measure represents a floor, not a ceiling.

Apply these measurement frameworks to your content

Technical fixes that must precede content changes

Technical plumbing forms the foundation for every other improvement. When the retrieval layer cannot read a page, every downstream investment in content, freshness, and query mapping lands on material the machine cannot access.

Three fixes come first.

1. Unblock AI crawlers. Check robots.txt for rules that block GPTBot, ClaudeBot, PerplexityBot, and Google-Extended. Blocking these crawlers is the most common silent reason a well-ranked page earns zero citations. This fix is a one-time correction with ongoing maintenance.
2. Add schema markup. Pages with proper schema markup are cited 2–4× more often in AI-generated answers than identical content without it. Priority schema types include Article, Organization, Person, FAQPage, and HowTo. Apply schema to every page, not only new ones.
3. Make pages machine-parseable. Answer-first formatting places the direct response within the first 40–60 words of each section. 44.2% of all LLM citations come from the first 30% of text on a page, which makes the opening paragraph the primary citation real estate. Every H2 should introduce a question that the following paragraph answers directly. Tables and FAQ blocks increase extractability.

For freshness signaling on pages where the URL is stable and cannot carry a year, use dateModified schema markup, lastmod in XML sitemaps, and HTTP Last-Modified response headers. These signals communicate recency to retrieval systems without requiring URL changes.

A neutral next-step checklist for the scissors pattern

The five items below form the minimum viable starting point for a content library that shows the impressions-up-clicks-down pattern. They are ordered by dependency so each item becomes true before the next one compounds.

1. Technical accessibility. AI crawlers are unblocked, schema appears on every priority page, and answer-first formatting is in place. Nothing downstream works without this.
2. Topic coverage audit. Map the fan-out question space for your three highest-impression, lowest-CTR query clusters. Identify gaps between what you have published and what the machine retrieves against.
3. Buyer-question alignment. Rewrite slugs, titles, H1s, and H2s on existing pages to match buyer language extracted from ChatGPT fan-out queries. Measure citation rate before and after.
4. Measurement readiness. Establish a baseline share-of-answer measurement across ChatGPT, Google AI Overviews, Perplexity, and Gemini before publishing anything new. Segment AI referrers in analytics. Set impression-decay tripwires in Search Console.
5. Refresh capacity. Confirm you have a mechanism to update priority pages on a continuous loop. In Arjun’s tests, a fixed library of any size decayed 78% to 99% in two months without maintenance. Volume and cadence stop being vanity metrics and become the entry fee for a game that resets weekly.

Walk through this checklist with your own data

Frequently asked questions about the scissors pattern

Is the impressions-up-clicks-down pattern always caused by AI Overviews, or are there other explanations?

AI Overviews are the dominant cause in 2026, but they are not the only one. Rich snippets, featured snippets, People Also Ask boxes, and Knowledge Panels all resolve queries on the SERP without requiring a click. Content decay, where a page loses citation share while holding its ranking position, produces the same scissors pattern in Search Console without any AI involvement. Intent misalignment, where a page ranks for queries it does not actually answer well, also drives impressions without clicks. The diagnostic approach stays the same regardless of cause. Compare impressions, clicks, CTR, and position across date ranges, filter by query type, and identify which query clusters show the sharpest CTR collapse. Informational queries beginning with “what,” “how,” and “why” are the most exposed to AI Overview suppression. Transactional and navigational queries are far less affected.

Do I need to stop doing traditional SEO and switch entirely to GEO?

No. The technical fundamentals of SEO, including site structure, crawlability, internal linking, page speed, and quality content, support both traditional search and AI citation. The optimization target and the success metric change instead. Traditional SEO aims for rankings on a human-readable list. GEO aims for citation inside a machine-generated answer. The authority model also differs. SEO earns authority through backlinks and domain authority, while GEO earns it through topical coverage across the full fan-out question space. On Arjun’s own site, articles structured for AI citation reached thousands of monthly Google impressions within weeks, and the GEO subfolder became the only source of new impressions on the domain. Relevant, structured, fresh, specific content performs on both surfaces. The measurement target is what needs to shift, from rank position to citations, mentions, and share of answer.

How do I measure share of answer without expensive tooling?

A manual baseline is achievable with no tooling cost. Build a fixed prompt set of 15 to 20 high-intent buyer questions drawn from sales call transcripts, support tickets, and comparison queries. Run each prompt across ChatGPT, Google AI Overviews, Perplexity, and Gemini. Record three outcomes per response: whether your brand was mentioned by name, whether a URL from your site was cited as a source, and whether the assistant actively recommended your business as the answer. Run each prompt five to ten times per engine to account for non-deterministic variance. Calculate share of answer as the number of responses citing your brand divided by the total tracked prompts, multiplied by 100. Re-measure every 30 to 60 days using the same fixed prompt set. Pair this with AI referrer segmentation in your analytics platform, where chatgpt.com and its equivalents appear as a distinct traffic class, and with impression and decay curve monitoring in Google Search Console. The combination gives you a measurable channel without requiring a dedicated monitoring platform to start.

How long does it take to see citations after restructuring content for AI retrieval?

Timelines vary by surface and by the type of change. Perplexity responds fastest because it favors content published or updated within the last 30 days. Google AI Overviews and ChatGPT take longer because their retrieval layers weight a broader set of signals. In Arjun’s documented tests on his own site, buyer-language relabelling of a jargon page produced citations within weeks of the change. Fan-out query alignment on rewritten pages produced citations while control pages remained uncited. New articles reached thousands of monthly Google impressions within weeks of publication. The GEO subfolder results mentioned earlier, where it went from zero to full impression dominance in 60 days, illustrate the compounding effect of topical authority. These findings come from Arjun’s own site and do not guarantee any specific outcome. Practitioner documentation generally shows initial citation signals in four to eight weeks, measurable share-of-answer improvement in one to three months, and compounding topical authority after month three.

What is the biggest mistake businesses make when they first see the scissors pattern in Search Console?

The most common mistake is treating the scissors as a content quality problem and responding with more content of the same type. As established earlier, this is a channel mechanics issue rather than a content quality issue, and the mistake lies in how businesses respond to it. Publishing more unstructured, uncrawlable, jargon-heavy content at the same cadence produces more of the same result. The second most common mistake is continuing to report on rank position as the primary success metric while the actual buyer journey has moved to AI-mediated discovery. A page can hold its ranking position while losing all of its citation share, and a standard rank report will still show everything as fine. The correct response to the scissors is to run the four-number diagnostic, identify which query clusters are most affected, fix technical plumbing first, map the fan-out question space, align content to buyer language, and shift the measurement target to citations and share of answer. The scissors pattern does not signal that SEO is broken. It signals that the channel changed and the measurement has not caught up.