USE CASES / TRAFFIC DECLINE
Traffic Declining Despite Stable Rankings?
“Our rankings are holding. Impressions are not collapsing. But sessions and clicks keep falling. What changed?”
That is a different problem from a normal ranking decline. When rankings fall, the diagnosis starts with lost positions, lost pages, new competitors, indexing changes, or a Google update.
When rankings remain stable but traffic falls, the click itself may be disappearing before a searcher reaches your site. AI Overviews can be part of that explanation, but they are not automatically the explanation.
The most credible response is to rule out ordinary causes before assigning the loss to AI. Seasonality, a core update, falling search demand, a tracking change, consent-mode changes, bot filtering, content cannibalization, SERP layout changes, and competitor activity can all reduce sessions while a narrow ranking view appears stable.
If those explanations do not account for the decline, then investigate whether the search result is answering more of the query before the click happens. This page helps you separate a measured traffic problem from an assumed AI explanation.
DEFINE THE SYMPTOM
State the symptom precisely
“Traffic is down but rankings are fine” is often too vague to investigate. Before deciding why traffic fell, define which metric fell and which metrics held. For example:
- Organic sessions fell, but Search Console clicks did not.
- Search Console clicks fell, but impressions remained stable.
- Click-through rate fell while average position stayed within the same range.
- A group of pages lost traffic while the overall domain did not.
- Non-brand traffic declined while brand traffic held.
- One country, device type, or landing-page group declined.
- A category of informational queries weakened while commercial pages held.
These are not interchangeable symptoms. Google Search Console measures Google search clicks, impressions, click-through rate, and average position. GA4 measures sessions and users subject to its implementation, consent, channel grouping, and filtering. A disagreement between the two can be a measurement question before it is a search-demand question.
Stable average position also does not mean stable SERP exposure. Average position can conceal movement across query variants, changes in the number of impressions per query, shifts in device mix, result-layout changes, and the difference between rankings on queries that still receive clicks and rankings on queries that do not.
The first job is to identify the missing click. The second is to explain it.
BEFORE YOU BLAME AI
Rule out the mundane causes first
This checklist is intentionally inconvenient. That is the point. A traffic-loss diagnosis that skips basic controls can turn an ordinary analytics or demand issue into an expensive AI search project.
01
Check seasonality and demand
Compare the affected period with the same period last year, not only the previous month. Look for seasonal demand patterns, budget-cycle changes, product or industry timing, holiday effects, regional events, changes in branded search demand, and a broad decline in impressions for affected query groups. If impressions are down materially, fewer searches may be happening. A stable ranking on a shrinking demand curve still produces less traffic.
02
Check whether the data changed before the search did
Did the GA4 implementation change? Consent mode, cookie controls, or tag-manager configuration? Channel grouping, referral exclusions, or bot filtering? Did a consent banner begin suppressing analytics events? Did a migration, redirect deployment, or domain change affect attribution? If GA4 sessions decline but Search Console clicks do not, investigate tracking and attribution before treating this as a search problem.
03
Check Google updates and technical change windows
Map the decline against known dates: core updates, spam updates, major SERP changes, site releases, template changes, internal-linking changes, robots.txt changes, canonical changes, noindex directives, redirect deployments, performance regressions, structured-data changes, and content releases or deletions. Do not claim an update caused the decline because the dates overlap. Use the overlap as a reason to inspect the pages and query groups affected.
04
Check cannibalization and content changes
A content push can reduce traffic without an obvious ranking collapse. Review whether you recently published overlapping pages, changed titles or headings, consolidated or split content, changed canonical targets, redirected old pages, removed supporting content, or replaced a high-performing page with a broader but less focused version. Two pages can trade impressions and positions while the overall picture appears stable.
05
Check the SERP layout before blaming AI
Search results change without AI Overviews being present. Inspect affected queries for paid ads, shopping units, local packs, video carousels, news modules, featured snippets, knowledge panels, forums, expanded People Also Ask results, and new competitors with stronger titles or rich-result treatments. A stable organic position beneath a larger result block can receive fewer clicks even when the ranking number looks unchanged.
06
Check competitor and market movement
Your rankings may be stable while a competitor earns more attention. Review whether competitors updated their title and snippet language, began appearing in rich results, earned a visible review, news, video, or forum result, published a new comparison resource, or became a more direct match for the query's current intent. Do not assume competitor activity because traffic fell. Look for the evidence in the affected SERPs.
THE EVIDENCE
What the evidence says about AI Overviews and clicks
Pew Research Center analyzed browsing data from 900 US adults who agreed to share online activity. The analysis covered 68,879 unique Google searches during March 2025, of which 12,593 produced an AI summary. The primary result is not that every AI Overview destroys traffic. It is that clicks differed meaningfully in this observed panel data.
| Observed behavior | Rate |
|---|---|
| Clicked a traditional search result, AI summary present | 8% of visits |
| Clicked a traditional search result, no AI summary | 15% of visits |
| Clicked a link inside the AI summary | 1% of visits |
| Ended the browsing session, AI summary present | 26% of pages |
| Ended the browsing session, no AI summary | 16% of pages |
The 8 percent versus 15 percent comparison is roughly a 47 percent relative difference in traditional-result clicks. It is not a 61 percent decline, and it should not be presented as one.
The 1 percent figure is equally important. Being cited in an AI summary may create visibility and credibility, but the study found direct clicks on those summary links were rare in the observed visits.
That changes the diagnostic. Your traffic may decline not because your ranking fell, but because a user received enough information to stop searching or because the click moved to a different source, result type, or answer surface.
Pew also found that 58 percent of panelists encountered at least one Google search result with an AI summary during the study period.
THE OTHER SIDE
Google's response to the study
Google publicly disputed the study's methodology on 22 July 2025, describing it as using a flawed methodology and a skewed queryset that is not representative of Search traffic, and stating that it continues to direct billions of clicks to websites daily without observing significant drops in aggregate web traffic.
That does not erase the Pew findings. It means this page should state the evidence and its limitation honestly.
Pew measured browsing behavior from a defined panel of 900 US adults during one month. It did not measure every Google user, every query class, or every industry. Google argued that the methodology did not reflect the full range of search behavior.
The appropriate conclusion is not that AI Overviews always cause a fixed traffic decline. It is this: in Pew's observed panel data, searches with AI summaries were associated with lower traditional-result click rates and higher session abandonment. That makes AI Overview exposure a plausible diagnostic factor when clicks fall despite apparently stable rankings. It does not establish universal causation for every site or query.
THE MISSING CLICK
What is different from a normal traffic drop
A conventional traffic drop often starts with less visibility. Your rankings fall, impressions fall, or your pages stop appearing for the queries that previously delivered clicks.
An AI-mediated traffic problem can look different. Your page may remain visible. It may retain its average position. It may even appear as a source in an AI answer. But the searcher may receive enough of the answer before the organic result to never visit your page.
The lost click is the critical distinction, and it matters because the response cannot be limited to recovering a ranking that did not fall. You need to ask which queries lost clicks, whether their impressions held steady, whether click-through rate declined, whether an AI Overview or other result treatment appeared, whether the query is now answered directly in the SERP, whether your page offers value beyond the answer shown, and whether your result is still the best next click after the searcher has seen the summary.
A page that only repeats a short factual answer is more exposed than a page that offers depth, original evidence, tools, comparison, decision support, implementation detail, or a useful next action. That is not an argument for padding pages. It is an argument for giving searchers a reason to continue.
DIRECTIONAL DIAGNOSTIC
Run a directional diagnostic in under 30 minutes
This diagnostic is triage. It can identify whether AI Overview exposure is worth investigating. It cannot prove that an AI Overview caused a traffic decline or establish your site-wide AI-search impact.
01
Choose four affected queries
Use queries that drove meaningful clicks before the decline, whose impressions are stable or close to stable, whose average position is stable or within a narrow range, whose click-through rate declined, and which represent a meaningful buyer need. Avoid building the diagnosis around a single vanity keyword.
02
Inspect the current SERP
For each query record the date, location and device, your organic position, whether an AI Overview appears, whether your page is cited, other major result blocks, competitors visible above or around your result, whether the answer can be completed without a click, and whether your title and snippet still offer a compelling next step. Use a consistent browser environment. Do not compare a logged-in personalized result with a neutral historical screenshot and call the difference a trend.
03
Compare the query before and after the decline
Use Search Console and your historical records to compare clicks, impressions, click-through rate, average position, landing page, device, country, and branded versus non-branded behavior. The goal is to identify a pattern, not to assign a cause from one data point.
04
Classify the likely situation
Demand issue: impressions fell, suggesting fewer available searches or lower visibility. Tracking issue: sessions and clicks disagree in a way that points to analytics or consent changes. SERP competition issue: rankings hold, but paid, local, video, shopping, competitor, or answer features changed the click opportunity. AI exposure hypothesis: rankings and impressions are broadly stable, click-through rate is down, AI Overviews appear for affected queries, and ordinary explanations do not account for the pattern. The fourth outcome is a reason to investigate, not proof.
WHAT IT PROVES
What the diagnostic can and cannot establish
A directional diagnostic can establish that clicks fell while rankings and impressions remained broadly stable and that answer-first SERP features are present. It cannot prove that an AI Overview caused the decline or guarantee that any change will restore prior traffic.
It can establish
- That clicks or sessions declined for a defined query and page group.
- That impressions and rankings remained stable enough to rule out a simple visibility explanation.
- That relevant SERPs contain AI Overviews or other answer-first features.
- That click-through rate changed after a result-layout change.
- That traffic loss may be concentrated by query type, page type, device, or location.
- That an AI-search investigation is justified.
It cannot establish
- That an AI Overview caused the decline.
- That every affected query has the same cause.
- That a citation inside an AI Overview will recover traffic.
- That content changes will restore prior click rates.
- That Google will show or remove an AI Overview on a fixed schedule.
- That an AI search program will recover a specific number of sessions.
DOCUMENTED ACTIONS
The response: documented actions and unproven claims
Start with the evidence you can verify.
Improve the page's post-answer value
If a result page answers the query directly, the next click must offer more than the answer alone. Strengthen useful depth: original data and methodology, first-hand expertise, a practical decision framework, product or implementation detail, comparative analysis, calculators or templates, clear buyer qualification, and current examples with source links. Google advises publishers to focus on helpful, reliable, people-first content, not content created to manipulate AI features.
Validate crawlability and indexation
Google says its AI features use the same underlying Search foundations. Important pages should remain accessible, indexable, canonicalized correctly, internally linked, and useful. Technical readiness does not guarantee traffic recovery. It removes avoidable blockers.
Measure AI visibility separately from organic traffic
Traffic, rankings, AI citations, and AI mentions are different metrics. A page can rank, be cited, and receive fewer clicks at the same time. Track organic clicks and click-through rate, query and landing-page exposure, AI Overview presence where observable, AI citations and mentions, competitor presence in AI answers, and the source domains used in relevant answers.
Strengthen source-oriented evidence where it serves buyers
The GEO research paper found that source-oriented changes, including authoritative references and citation-aware content, can improve visibility in generative-engine responses. It does not guarantee traffic, citations, rankings, or revenue for an individual page.
What is not documented
Do not promise that:
- Adding llms.txt will recover clicks from Google AI Overviews.
- Special schema will make Google remove an AI Overview.
- A citation inside an AI Overview will restore former organic click-through rate.
- Publishing more AI-targeted pages will reverse a traffic decline.
- ChatGPT crawler access controls Google AI Overview behavior.
- A recovery will happen within a fixed number of weeks.
Google explicitly says it does not require machine-readable AI text files, special markup, or special schema for its generative AI search features. For ChatGPT's own search surfaces, OpenAI documents OAI-SearchBot as the crawler used to surface sites. It does not document a fixed route from crawler access to referral traffic, or a correction mechanism for Google behavior.
The correct response depends on the diagnosis. Do not apply an AI-search remedy to a demand, tracking, technical, or conventional SERP-competition problem.
WHEN IT IS NOT AI
When this is not an AI problem
Do not treat every traffic decline as an AI Overview issue. This is not an AI-search engagement when:
Search demand fell
Impressions declined across the affected query set, particularly in a year-over-year comparison.
Analytics changed
GA4 sessions and Search Console clicks disagree in a way that points to consent, tagging, attribution, or bot-filtering changes.
Rankings or eligibility declined
The pages lost position, impressions, indexability, or key SERP features.
A conventional SERP feature displaced clicks
Paid ads, local results, shopping units, video, news, featured snippets, forums, or another result type changed the click opportunity.
Content architecture changed
Cannibalization, redirects, canonical changes, removed pages, or internal-linking changes explain the loss.
A competitor created a clearer result
The response is standard SEO, content, product, or positioning work.
The affected queries are not commercially relevant
Recovering traffic that does not support the business is not a useful objective.
There is no evidence beyond one traffic chart
A correlation between an AI Overview and lower traffic does not establish causation.
The expected outcome is guaranteed recovery
No credible program can promise that Google will remove an AI Overview or restore a prior click-through rate by a specific date.
The goal is not to make AI the explanation. It is to find the explanation that survives inspection. If the ordinary checks account for the decline, fix that problem. If they do not, and a defined query set shows stable visibility with falling clicks alongside answer-first SERP exposure, then an AI-search investigation is justified.
TAKE THIS INTERNALLY
A decision-ready brief for the traffic problem
Before assigning work or engaging an external partner, write the problem down in a form that can be tested.
The affected period
State the start and end date, plus the comparison period.
The affected metric
Specify whether the decline is in Search Console clicks, GA4 organic sessions, conversions, revenue, or another metric. Do not call them interchangeable.
The affected scope
List the affected query groups, landing pages, devices, countries, and branded or non-branded segments.
What remained stable
Document impressions, average position, rankings, or other signals that did not move materially.
What changed in the SERP
Record AI Overviews, ads, shopping, local results, featured snippets, competitor movement, and other visible result changes for the representative query set.
What mundane causes were ruled out
Note seasonality, demand, tracking changes, consent changes, releases, technical changes, cannibalization, and Google-update timing.
The AI exposure hypothesis
State it as a hypothesis, not a conclusion: for this defined query set, clicks fell while impressions and average position remained broadly stable, answer-first result features appeared on representative SERPs, and the ordinary causes reviewed so far do not explain the decline.
The next measurement step
Define what will be monitored: query-level clicks, click-through rate, impressions, position, landing-page behavior, AI Overview exposure where observable, cited sources, competitor presence, and commercial outcomes.
This keeps the work grounded in evidence rather than panic.
WHERE THIS PAGE FITS
Where this page fits
This page addresses a specific traffic symptom: rankings and possibly impressions appear stable, but organic clicks or sessions decline.
- For the complete scenario library, see Use Cases.
- For measuring brand mentions, citations, position, and source composition in AI answers, see AI Visibility.
- For the technical and content foundations relevant to ChatGPT search surfaces, see ChatGPT SEO.
- For the practitioner framework that separates durable SEO work from AI-search additions, see LLM SEO.
- For a competitor-presence problem in ChatGPT, see Competitor Showing Up in ChatGPT.
- For inaccurate information in AI-generated answers, see Fix AI Hallucinations About Your Brand.
- Measuring AI search ROI remains a separate scenario page and is not yet live.
FAQ
Frequently asked questions about falling traffic with stable rankings
Why is organic traffic falling when our rankings are stable?
Stable rankings do not guarantee stable traffic. Search demand may have declined, the SERP layout may have changed, your click-through rate may have fallen, competitors may have gained more visible result treatments, or analytics implementation may have changed. AI Overviews are one possible factor, but they should be investigated only after ordinary demand, tracking, ranking, technical, and SERP explanations have been reviewed.
Can AI Overviews reduce organic clicks without changing rankings?
They can be a plausible factor. Pew Research Center found that users clicked a traditional search result in 8% of visits with an AI summary, compared with 15% of visits without one. The study observed an association in a defined panel, not universal causation for every site or query. A traffic diagnosis should therefore review affected queries, click-through rate, impressions, SERP layout, and other explanations before attributing a decline to AI Overviews.
What did the Pew Research Center study find about AI Overviews and clicks?
Pew Research Center analyzed 68,879 unique Google searches from 900 US adults during March 2025. Traditional search-result clicks occurred in 8% of visits with an AI summary and 15% of visits without one. Clicks on links inside AI summaries occurred in 1% of visits with a summary. Users ended their browsing session after 26% of visits with an AI summary, compared with 16% of visits without one.
Did Google dispute the Pew Research Center AI Overview study?
Yes. Google publicly disputed the study's methodology after publication, describing it as using a flawed methodology and a skewed queryset not representative of Search traffic. The study remains useful observed panel evidence, but it should not be treated as a universal estimate of click behavior across every query, industry, device, or Google user. The appropriate use is as a reason to investigate AI Overview exposure in a defined traffic-loss diagnosis, not as proof of causation.
What should I check before blaming AI Overviews for a traffic decline?
Check seasonality and search demand, GA4 and consent-mode changes, Search Console reporting consistency, Google updates, site releases, crawlability and indexation, redirects and canonicalization, content cannibalization, SERP features, competitor movement, and device or country differences. If one of these explains the decline, address that cause before beginning AI-search work.
How can I test whether AI Overviews are affecting our clicks?
Select four commercially meaningful queries where clicks and click-through rate declined while impressions and average position remained broadly stable. Inspect the current SERP for AI Overviews and other answer-first features, document your position and visible result blocks, then compare query-level clicks, impressions, click-through rate, and position before and after the decline. This is directional triage, not proof that an AI Overview caused the loss.
Does being cited in an AI Overview guarantee referral traffic?
No. Pew Research Center found that users clicked links inside AI summaries in 1% of observed visits with a summary. A citation may still support visibility or buyer awareness, but it does not guarantee clicks, qualified traffic, leads, or revenue. Measure citations and traffic separately.
Can improving content recover traffic lost to answer-first search results?
It can improve the value of a click, but no content change guarantees traffic recovery. Pages should provide useful value beyond a short answer, such as original research, practical tools, implementation guidance, comparisons, first-hand expertise, or decision support. Google recommends helpful, reliable, people-first content rather than content created to manipulate AI features.
Will llms.txt or special schema markup recover traffic from Google AI Overviews?
Google states that it does not require machine-readable AI text files, special markup, or special schema for its generative AI search features. These are not documented methods for recovering clicks from AI Overviews. Maintain accurate structured data where it faithfully represents visible content, but do not present it as a guaranteed traffic-recovery mechanism.
When is a traffic decline not an AI-search problem?
It is not primarily an AI-search problem when falling search demand, analytics changes, consent configuration, tracking errors, lost rankings, indexation problems, content cannibalization, conventional SERP features, technical regressions, or competitor movement explain the loss. It is also not an appropriate AI-search engagement when the affected queries are not commercially relevant or leadership expects a guaranteed traffic recovery.
Rule out the ordinary first. Then bring us the gap.
If you have worked through the mundane causes and still have a defined query set losing clicks without a conventional visibility explanation, an AI Search Assessment establishes where your brand appears in AI answers, which sources those platforms use, and which competitors appear instead. It will not promise that Google removes an AI Overview or that a prior click-through rate returns.