AI Search

AIO Erosion (Suspected)

A per-query heuristic for AI Overviews impact. Google Search Console does not let us combine its searchAppearance dimension with the query dimension, so we cannot prove per-query AIO presence from GSC alone. Instead, we surface queries with the classic AIO fingerprint and only fire the report when site-level AIO impressions actually rose in the window.

Data availability. This report depends on Google returning an AI Overviews value in Search Console's searchAppearance dimension. Google added that bucket in 2026 and coverage is still rolling out; as of September 2026 most properties — including sites with heavy AI Overview exposure — get no AIO rows from the API at all. When Google returns nothing, this report shows a "no AI Overviews impressions detected" notice rather than zeros, and no "estimated clicks lost" figure is computed. Treat it as a monitor that switches on when Google's data does, not as a number you can rely on today.

When the report fires

Two conditions must both be true:

  1. Site-level AIO ramp. The window is split in half; the second half must either start from zero AIO impressions or grow by ≥25% versus the first half. Otherwise we don't have enough signal to attribute per-query CTR changes to AIO, so the report shows a notice instead of rows.
  2. Per-query pattern match. The query must show: impressions ±30% (held steady), and CTR drop ≥30% (relative) vs the immediately preceding equal-length window.

How we compute it

  1. Aggregate every query in the current and prior equal-length windows from gsc_query_day.
  2. Inner-join on query; keep rows that clear your site's impression floor in both windows.
  3. Keep rows where impressions held (within ±30%) but CTR fell by ≥30% relative.
  4. Compute estimated clicks lost as max(0, prior_ctr − current_ctr) × current_impressions.
  5. Sort by estimated clicks lost, descending.

How to read each row

  • Query — the search term whose CTR collapsed.
  • Impressions — current-window total. Stable vs prior is the whole point.
  • Prior CTR / Current CTR — the before-and-after that triggered the row.
  • CTR Drop — relative drop, negative because CTR fell.
  • Estimated Clicks Lost — the upper bound on impact for this query in this window. Sort by this column to triage by upside.
  • Avg Position — your impression-weighted position. If you're still ranking the same, a snippet rewrite has the best shot of clawing back clicks.

What to do with it

  1. Open the top 5 queries in a private browser session and check the SERP — verify AIO is actually present.
  2. Find the page ranking for each query. Rewrite the title and meta description with a differentiating angle that AIO summaries don't capture: a fresh date, a contrarian take, a specific number, a guarantee, a brand voice.
  3. Add structured FAQ or HowTo schema where the intent matches — that's how some publishers earn placement inside the AIO and convert it back into clicks.
  4. Re-check after 2–3 weeks. Rows that drop off the list are wins.

Caveats & limits

  • Heuristic, not proof. A query can fit the fingerprint for other reasons — a new competitor, a SERP layout change, seasonal intent shift. Treat rows as strong candidates that need a human SERP check.
  • Queries with very low impressions are filtered out, using a floor scaled to your site rather than a fixed number (a fixed 50 in both windows locked smaller sites out of this report entirely). For sites where your top queries sit below this threshold, widen the date range.
  • The 30% CTR-drop threshold is conservative. Real-world AIO erosion can be subtler; you'll catch the worst offenders here, not every affected query.

Related reports