What should a team validate first when tracking ai overview impressions in google search console: the 2026 reality appears?

Reproduce the problem in the live implementation, isolate whether it is scoped to one report or flow, and compare it against at least one secondary source before changing the setup.

How do I know whether the fix actually worked?

You need before-and-after evidence in the browser and in the downstream report. A clean-looking dashboard without validation is not enough.

When should this become a full GA4 audit instead of a quick fix?

If the issue touches attribution, consent, revenue, campaign quality, or data trust for more than one workflow, it is usually safer to audit the surrounding implementation than patch only the visible symptom.

Tracking AI Overview Impressions in Google Search Console: The 2026 Reality

Can I see AI Overview impressions in Search Console?

No — Search Console does not have a native AI Overview filter, and Google has confirmed none is planned. AI Overview impressions, AI Mode impressions, featured snippets, and standard 10-blue-link impressions are all blended together under the "Web" search type with no way to separate them.

This was confirmed by Google's John Mueller in September 2025 after a fake-filter screenshot circulated, and reaffirmed in Google's official documentation in 2026. The only ways to estimate AIO performance today are heuristic methods (query-pattern regex filters in GSC) and third-party tools that combine GSC data with their own AI-Overview tracking (Semrush, SISTRIX, Ahrefs, AlsoAsked).

This post explains the four documented Search Console discontinuities since May 2025, the impression-inflation bug Google quietly disclosed in April 2026, and the practical reporting framework that actually works in this environment.

What Google's documentation actually says

The official position, updated in Search Console help in late 2025: *"Just like the rest of the search results page, sites appearing in AI features (such as AI Overviews and AI Mode) are included in the overall search traffic in Search Console."*

Translation: there is no filter, and there will not be one in the immediate future. Here's what counts as what:

Surface	Click counted as	Impression counted as	Position
AI Overview citation	A click in GSC	Impression only when scrolled or expanded into view	Single position; all AIO links share that position
AI Mode link	A click in GSC	If your page appears in the AI response	Each component (link card, image, carousel) has its own position
AI Mode follow-up question	New query	New query	Reset to position 1
Featured snippet	A click	Impression	Position 1 (typically)
Standard organic	A click	Impression	Calculated normally

A URL appearing in both an AI Overview AND a blue-link organic result for the same query is counted as a single impression — Google does not double-count. This matters because it means impression *growth* in 2025–26 reflects genuine visibility changes, not duplicate counting artefacts.

The four discontinuities that broke YoY comparisons

Search Console didn't quietly degrade once between May 2025 and April 2026 — it degraded on at least four documented dates. Annotate every one of these on your YoY dashboards:

13 May 2025 — Impression-inflation bug begins

Google over-reported impressions starting on this date. The bug was disclosed in a 47-word changelog entry on the Data Anomalies page on 3 April 2026 — no email, no Search Liaison post, just a quiet update. For roughly 11 months, impression data was inflated by an estimated 30–50% across 17 analysed properties (Passionfruit research, April 2026).

17 June 2025 — AI Mode merges into "Web"

Google began counting AI Mode clicks, impressions, and position toward Search Console "Web" totals with no native filter to separate them. AI Mode now reaches roughly 75 million daily users in the US. Every metric calculated from "Web" totals after this date includes AI Mode contribution.

12 September 2025 — &num=100 parameter killed

Google removed the URL parameter that let users (and SEO tools) view 100 results per page. Rank trackers and analytics tools that relied on it lost a long-standing signal. Reported impression patterns shifted again as a side effect — many properties saw a sudden impression drop on this date that wasn't a traffic change at all, just removal of scraper-inflated impressions.

3 April 2026 — Inflation correction begins rolling out

Google's April 3 changelog announced the rollout of corrected impression counts. Properties will see impressions drop after this date — not because traffic changed, but because the inflation finally stopped. The correction rolled out over several weeks. Clicks were not affected (Google's confirmation).

The implication for any agency or in-house team: YoY comparisons spanning these four dates are not meaningful for impression metrics. Pivot to clicks as your primary KPI. Treat impressions as directional only until June 2026 establishes a new clean baseline.

What you can do today: the heuristic filter approach

Since there's no native filter, the working alternative is to identify queries that *likely* triggered AI Overviews based on query characteristics. The Otterly.AI / Rankfuse approach is the most practical:

Step 1 — Open GSC Performance, last 28 days, Queries view

Filter by Query → contains. You're going to apply a regex pattern that matches conversational, AI-style queries — the kind that overwhelmingly trigger AI Overviews.

Want to see whether attribution loss is already distorting your channel data?

Audit attribution

Step 2 — Apply this regex (custom regex filter, "Matches regex")

This catches longer-form, conversational queries containing question words, comparison signals, or temporal/explanatory qualifiers. They're proxy for AI Overview eligibility.

Step 3 — Compare CTR vs your full-property baseline

If your filtered (AI-likely) CTR is meaningfully lower than your full-property average — and especially if it's around 0.6% vs your 1.7% baseline — you're seeing the AIO-cannibalisation effect. Seer Interactive's Q4 2025 study found organic CTR dropped from 1.76% to 0.61% (61% decline) for queries with AI Overviews, while paid CTR dropped 68% on the same queries (from 19.7% to 6.34%).

Step 4 — Drill into Pages tab (filter still applied)

The pages with high impressions but very low CTR under this filter are your most AIO-vulnerable pages. They're likely being cited or summarised in AI Overviews — readers see the summary and don't need to click. These are the pages to prioritise for refresh, FAQ schema, and answer-first restructuring.

Step 5 — Drill into Countries (filter still applied)

AI Overviews roll out at different times in different regions. Watching CTR by country with the AI-prompt filter on lets you see when AIO becomes dominant in a new region — typically impressions stay stable while clicks drop sharply in a specific country first.

The third-party tool approach — more accurate, paid

For properties where AIO performance is critical, combine GSC with a tool that tracks which queries trigger AI Overviews:

Export your GSC top-query data for a defined comparison period (e.g. last 30 days vs prior 30 days). Include clicks, impressions, CTR, average position.
Export an AI Overview visibility report from a third-party tool — Semrush, SISTRIX, Ahrefs, AlsoAsked, and most rank trackers now flag which keywords show AI Overviews in the SERP.
Merge in a spreadsheet on the query column. You now have GSC performance data joined to AIO presence per query.
Calculate the change in performance per query. Separate queries into two groups: gained AIO visibility during the period, vs lost AIO visibility during the period. Compare metric changes between the groups.

Pages that lost AIO visibility but kept impressions stable are competitors. Pages that gained AIO visibility but saw click drops are textbook "Great Decoupling" cases.

"The Great Decoupling" — what stakeholders actually need to know

If your impressions are growing while clicks stagnate or decline, you're seeing the pattern industry observers have called the Great Decoupling. The combination of factors:

AI Overviews are satisfying user intent on the SERP — the user gets the answer without clicking.
CTR drops disproportionately at the top positions. Ahrefs' analysis found AI Overview presence is associated with about 34.5% lower average CTR for the #1 organic result on affected queries (Ahrefs, 2025).
Citations within AI Overviews are formatted less prominently than traditional organic results — multiple citations share the same position, dividing potential clicks.
Users may read the AI-generated answer without examining citations at all — you got the impression but the click never happened.

For stakeholder reporting, this means traditional "impressions are up, we're winning" framing is dangerous in 2026. The healthier frame:

Lead with clicks, not impressions. Google itself confirmed clicks were not affected by the impression bug — clicks are your most defensible metric.
Tie clicks to business outcomes wherever possible. Conversions, demos booked, form fills, revenue per landing page — these are the metrics the impression bug couldn't touch.
Annotate the four discontinuity dates on every YoY chart. Don't let stakeholders compare May 2025 impressions to April 2026 impressions without context.
Establish a new clean baseline starting June 2026. Genuine YoY comparisons become possible only from June 2027 onwards.

Setting realistic expectations: the "AI Tax" calculation

Different sites lose different percentages of clicks to AIO. The AI Tax (RankFuse's term) is the percentage of click-through-rate you've lost on AI-eligible queries:

Apply your AI-prompt regex filter in GSC.
Compare the CTR of this segment from the start of your 16-month available data window (when AIO was less prevalent) vs the last 3 months.
The percentage drop is your AI Tax on informational queries.

Industry benchmarks for context:

Ahrefs: ~34.5% CTR drop on top organic positions when AIO present
Search Engine Land/Seer: 61% drop in organic CTR for queries with AIO
Site-specific AI Tax we've measured in audited properties: 25–55% range, depending on content mix

A B2B SaaS with predominantly informational queries will see a higher AI Tax than an e-commerce site with predominantly transactional queries.

Important caveat: GSC's API only retains 16 months of data. As we move deeper into 2026, the pre-AI baseline data from late 2023/early 2024 is falling off the rolling window. Export your historical data now if you haven't — once it's gone you cannot prove a traffic drop was a platform shift rather than a performance failure.