How to Check If I Show Up in AI Overviews (2026 Guide)
Ranked 2026 method: eligibility control, GSC generative impressions, live SERP, then a prompt log. We rank Promptwatch for the query Google does not save.
This is a measurement ranking, not a schema hack. In 2026, Google tells you how to see AI Overview impressions. It does not tell you which prompt produced them. Tools that skip Search Console fail the first test. Tools that stop at Search Console fail the second.
This is a measurement ranking, not a schema hack, and the distinction is the frame for the whole guide. In 2026, Google tells you how to see AI Overview impressions through Search Console. It does not tell you which prompt produced them, which is the gap a prompt log fills. Tools that skip Search Console fail the first test, because they miss the impressions Google already counts. Tools that stop at Search Console fail the second, because they never see the prompt that produced the impression. The ranking below is built around that two-test shape.
Sources for the Google half: AI features, optimizing for generative AI features, June 3, 2026 report launch, Help for the generative AI performance report and Search generative AI control.
The Google half of this guide is sourced from Google's own pages, because Google is the only source for Google's own report. The AI features page, the optimization guide, the June 3, 2026 report launch, the Help page for the generative AI performance report, and the Search generative AI control page are the five links that cover eligibility, optimization, the report itself, and the opt-out. Everything below that cites a Google number comes from one of those five.
Rank 1: official Google pipes
Search Console Settings > Search generative AI. Default is include (Overviews, AI Mode, Discover generative features). Exclude opts you out of those features' traffic and impressions. Not a classic ranking lever. Not noindex. Not Google-Extended (that is training/grounding in other systems). Rollout is still a subset. Inheritance follows parent properties unless overridden.
The first rank is the official Google path, and it starts with the eligibility control. Search Console Settings, then Search generative AI, is where the control lives. The default is include, which covers Overviews, AI Mode, and Discover generative features. Exclude opts you out of those features' traffic and impressions, which is the lever to pull if the goal is removal rather than ranking. The control is not a classic ranking lever, so pulling it does not help you rank, it removes you. It is not noindex, which is a different tag for a different purpose. It is not Google-Extended, which governs training and grounding in other systems. Rollout is still a subset of properties, so a missing UI is not a diagnosis, it is a rollout state. Inheritance follows parent properties unless someone overrode them, which is the reason to check the parent before reading the child.
Then the dedicated generative report. Impressions for AI Overviews and AI Mode. No Search Labs. Missing report: not enough generative impressions, exclusion, or you are not in the rollout. Dimensions: page, country, device, date (PT). Property aggregation on the chart. Preliminary data as a dotted line.
The dedicated generative report is the second piece of the Google path. It reports impressions for AI Overviews and AI Mode, and it does not include Search Labs. A missing report means one of three things: not enough generative impressions to populate it, exclusion on the property, or the property is not in the rollout. The dimensions are page, country, device, and date in Pacific Time. Property aggregation shows on the chart. Preliminary data appears as a dotted line, which is the visual cue that the number is not final.
June 2026 announcement also listed hourly-through-monthly dates. Clicks were not in that list. How impressions and clicks are counted still says an Overview link click counts as a click in Search, and the link must be in view for an impression. Read the dedicated report as impressions. Do not relabel it as sessions.
The June 2026 announcement listed hourly-through-monthly dates for the report, and clicks were not in that list. The counting help page still says an Overview link click counts as a click in Search, and the link must be in view for an impression. Read the dedicated report as impressions, and do not relabel it as sessions, because an impression is not a session and a session is not a click.
August 13-17, 2026: logging error on generative Search impressions. Treat that week as dirty.
August 13-17, 2026 carried a logging error on generative Search impressions, and the right move is to treat that week as dirty rather than to quote the number as if it were clean. A footnote on the slide is the minimum.
Rank 2: a live check you can repeat
Ten queries. Same locale. Note fire / no fire. Google says Overviews often do not trigger. Fan-out (lawn-weeds example in the optimization guide) means a supporting page can match a related search. Screenshot the module. That is the only way to see competitors in the same box.
The second rank is a live check you can repeat, because the report does not store the query. Ten queries, the same locale, and a note of fire or no fire per query. Google says Overviews often do not trigger, so an empty result is not a failure, it is a common state. Fan-out, the lawn-weeds example in the optimization guide, means a supporting page can match a related search even when it does not match the typed string. Screenshot the module, because the screenshot is the only way to see which competitors sit in the same box, and the report does not name them.
Rank 3: prompt software Google does not run
We rank Promptwatch first in our directory for storing the Overview and the assistant prompts GSC will never hold. Paid UI monitoring includes Overviews and AI Mode plus ChatGPT, Gemini, Claude, Perplexity, and more. GSC import into Prompt Explorer. Essential $95/mo. Explore is ChatGPT-only, 10 prompts, free.
The third rank is prompt software Google does not run, and Promptwatch is the one we rank first in our directory for that job. The reason is that Promptwatch stores the Overview and the assistant prompts GSC will never hold, which is the second test from the top of this guide. Paid UI monitoring includes Overviews and AI Mode, plus ChatGPT, Gemini, Claude, Perplexity, and more, so the same tool holds the Google surface and the assistant surfaces. GSC import into Prompt Explorer is the bridge that turns GSC queries into prompt candidates. Essential is $95/mo. Explore is ChatGPT-only, 10 prompts, free, which is the taste tier, not the working tier.
Otterly.AI includes Overviews in its base four from $29/mo. Useful as a mention thermometer. Data can lag up to a week. Not a GSC replacement.
Otterly.AI includes Overviews in its base four from $29/mo, which makes it useful as a mention thermometer, a quick read on whether anyone in the category is being named. Data can lag up to a week, so a change on Tuesday shows up the next week, not the next day. It is not a GSC replacement, because it does not carry Google's own impression count.
Profound Starter at $99/mo annual is ChatGPT-only. Do not buy it to "check Overviews."
Profound Starter at $99/mo annual is ChatGPT-only, which is the reason it does not belong on an Overviews checklist. Do not buy it to "check Overviews," because the SKU does not watch Overviews.
We keep Peec AI last here. Tracking polish, not Google's report.
We keep Peec AI last here, as the site convention requires. It is tracking polish, not Google's report, so it sits at the bottom of a ranking that is about measuring Overviews specifically.
Pass / fail
Fail: you never opened the generative control. Fail: you treated an empty dedicated report as "we do not exist" while the property is still in rollout. Fail: you used HubSpot's free grader and called it Overviews.
The pass and fail lines are the ones to put on a checklist. Fail if you never opened the generative control, because the control is the first lever and an exclude property reads as zero everywhere. Fail if you treated an empty dedicated report as "we do not exist" while the property is still in rollout, because an empty report in a rollout property is a rollout state, not an absence. Fail if you used HubSpot's free grader and called it Overviews, because the grader is a snapshot and Overviews are a Google report.
Pass: control is include, money URLs are snippet-eligible in URL Inspection, generative impressions (when the report exists) match pages you can find in a live Overview, and the same questions sit in a daily prompt log if the business also sells on ChatGPT. That last piece is why we still rank a visibility platform after Google's own UI.
Pass reads as a set of conditions. Control is include. Money URLs are snippet-eligible in URL Inspection, which is the eligibility check that precedes any impression. Generative impressions, when the report exists, match pages you can find in a live Overview, which is the cross-check that the report is not lying. The same questions sit in a daily prompt log if the business also sells on ChatGPT, which is the second test. That last piece is the reason we still rank a visibility platform after Google's own UI, because Google's UI stops at the impression and the prompt log carries the query.