Diagnostic · Six Checks
An AI Overview loss and an algorithm penalty look identical in a sessions chart and are diagnosed completely differently. Six checks separate them in about an hour.
By Izzy Gregorio · Updated August 2026 · 10 min read
In short
An AI Overview traffic loss and an algorithm penalty look identical in a sessions chart and are diagnosed differently. The tell is impressions holding steady while clicks fall, concentrated on informational queries. Six checks separate the two in about an hour. Some of that traffic is not coming back, and the correct response is competing for the citation instead.
Before you start
Google Search Console access, GA4 access, and about an hour. Before you run the checks, it helps to know the size of the effect you are looking for.
organic click-through where an AI Overview appears, from 1.76% to 0.61%
organic click-through where no AI Overview appears at all
click-through for brands cited inside the Overview, against uncited brands on the same page
Seer Interactive, September and November 2025. 3,119 informational queries, 42 organizations, 25.1 million organic impressions.
The middle number complicates the diagnosis, and you should know that going in. Clicks are falling 41 percent even on queries where no AI Overview appears, which means the impressions-steady-clicks-down signature is directional rather than definitive.
The third number is why any of this is worth fixing. Winning the citation slot is not a consolation prize. It is measurably the difference between two brands on the same results page.
Check 1
This is the single most diagnostic comparison available and most reports never run it.
Open Google Search Console, go to Performance, and set a date comparison covering the drop window against the same length period before it. Enable both the Clicks and Impressions metrics.
Impressions steady or up, clicks down. You are still surfacing for the same queries and fewer people are clicking. That pattern points to an answer being served above your result. This is the AI Overview signature.
Impressions down alongside clicks. You lost visibility itself, not just the click. That is a ranking or indexing problem and it is diagnosed differently.
Write down which one you are looking at before doing anything else. Everything downstream depends on it.
Check 2
Filter your Search Console query report by question words: what, how, why, when, is, does, can.
AI answers absorb informational demand first, because a definitional question is fully answerable in the summary. A buyer researching how something works never needed your page, they needed the answer. Commercial and comparison queries hold up better, because buyers still click through to evaluate specifics, pricing, and proof.
If your losses cluster in the question-word set while your commercial terms are flat, that is corroboration for an AI Overview loss and it also tells you which pages to restructure first.
Check 3
Pull the drop start date and compare it against Google's published Search Status update history.
A drop that begins the day a confirmed core update starts rolling out is an algorithm event until proven otherwise. A drop with no corresponding update, on steady impressions, in informational queries, is the AI Overview pattern.
Both can be true at once. Do not stop at the first plausible explanation, which is the most common mistake in this diagnosis.
Check 4
Take your ten highest-loss queries from Check 2 and search each one directly, in a logged-out browser window.
For each query, record three things: whether an AI Overview appears, whether your domain is cited inside it, and which domains are cited instead.
That third column is the one that matters. If an AI Overview is present and your competitor is cited inside it while you rank below it, you did not lose a ranking. You lost a citation slot, and no amount of ranking work recovers it.
That competitor is also currently earning roughly 35 percent more clicks than you from the same results page, per the Seer data above. Same query, same page, one of you named.
Check 5
In GA4, check the dedicated AI Assistant channel, which Google added in July 2026 specifically to separate this traffic. If your property predates it or you want a cross-check, build a session source segment for the AI platform referrers: chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, copilot.microsoft.com.
Compare that segment across the same window as the drop. Rising AI referrals during falling organic sessions means demand did not disappear, it relocated. That is a very different internal conversation than lost demand, and it reframes the budget question from recovery to redistribution.
Note the limitation honestly. AI referral traffic is a lagging indicator and undercounts badly, because a buyer who reads about you in an answer and later types your name arrives as direct or branded search. Treat it as a floor, never a full measure.
Check 6
Open your three highest-loss pages and read the first hundred words of each.
Ask one question: if a model lifted the opening passage into an answer with no surrounding context, would it still make sense and would it name your brand?
Most pages fail this. They open with a runway paragraph about how the industry is changing, then reach a conclusion three sections down. That structure is invisible to extraction, because there is no self-contained passage to lift.
The fix is editing, not rewriting. Move the answer to the top, make it 40 to 60 words, name the entity explicitly, and remove any backward references. Then give every section its own direct answer before elaboration.
Expectations
Some of this traffic is gone permanently and you should plan around that rather than chase it.
The behavior shift underneath is not temporary. G2's Answer Economy survey of 1,076 business-to-business buyers in March 2026 found 51 percent now start research in an AI chatbot rather than a search engine, up from 29 percent eleven months earlier, and 53 percent find AI research more productive than traditional search, up from 36 percent seven months prior.
Buyers did not stop researching. They stopped needing a results page to do it.
What is recoverable is the citation. When your competitor is named inside the AI Overview appearing above your listing, that slot is contestable, and it is worth roughly 35 percent more clicks than not being named. Google AI Overviews favor content already performing in Google search, which means your existing rankings are an asset in that fight, not a consolation prize.
What is not recoverable is the informational click that used to arrive because the answer lived only on your page. That answer now lives in the summary. Optimizing that page for clicks is chasing a category of demand that has moved.
Keep going
The next shift will look like this one: a chart that drops, an explanation that arrives late, and a report that measured the wrong thing. Monthly, what moved in AI citations, which engines changed, and the checks worth running.
No spam. Unsubscribe anytime.
Measurement
Sessions stopped being a complete picture the moment answers started resolving above the results. Add three gauges.
| Metric | What it tells you |
|---|---|
| Share of model | How often you appear in category answers across engines |
| Citation rate | How often those mentions actually link to you |
| Competitive citation gap | Your share of model minus your closest competitor's |
Report monthly. AI content freshness cycles run on roughly a 70-day cadence, so a quarterly report describes a position that has already changed twice.
Common questions
Compare clicks against impressions in Google Search Console over the drop window. Impressions holding steady while clicks fall, concentrated in informational queries, is the AI Overview signature. Impressions falling alongside clicks indicates a ranking or indexing problem instead. Note that Seer Interactive found clicks fell 41 percent even on queries without an AI Overview, so the signature is directional rather than definitive.
Partly. Informational clicks that AI answers now resolve directly are unlikely to return, because the behavior change is durable. Citation slots inside AI Overviews are contestable and worth roughly 35 percent more clicks than not being cited, per Seer Interactive. Commercial and comparison queries hold up better since buyers still click through to evaluate specifics.
Blocking removes you from AI answers entirely without restoring the clicks, because the answer gets assembled from other sources instead. You lose the citation and the referral while a competitor gains both. Note that OpenAI's crawlers are separately controllable: blocking GPTBot withholds training data while allowing OAI-SearchBot preserves citation eligibility.
Definitional and how-it-works pages lose first, since a summary can fully answer those questions. Pricing, comparison, case study, and product pages hold up better because buyers want specifics, proof, and options they can evaluate themselves before deciding.
Not necessarily, and blaming the agency often misdiagnoses it. The relevant question is whether they measured the right thing. BrightEdge found only around 17 percent of AI Overview citations come from pages ranking in Google's organic top ten as of February 2026, so a rankings report cannot detect this loss.
It depends on which fix applies. Structural edits to a page move fastest, since they only need the page recrawled. Crawler access problems resolve within a crawl cycle. Citation supply work takes longest because it depends on third parties publishing on their own schedule, and nobody controls that directly.
After the six checks
Check 4 tells you which competitor is named on ten queries. The AI visibility audit from Conspicuouz Creative Group runs the same test at scale across ChatGPT, Claude, Gemini, and Perplexity, records every competitor named in your place, and identifies which of the structural causes is producing the pattern. You get the findings whether or not you ever work with us.