AI Search

AIO-Resistant Pages

The inverse of AIO Erosion. Surfaces the pages whose CTR held or improved through the same window in which AI Overviews coverage ramped on your site. These are the templates that AIO didn't eat — study them and copy the patterns across to pages that are losing clicks.

When the report fires

The "resistant" framing only makes sense when AI Overviews is actively rising on your site. The report uses the same site-level ramp gate as AIO Erosion: the second half of the window must show ≥25% more AIO impressions than the first half (or start from zero). If AIO didn't ramp, the report shows a notice instead of rows — saying "your CTR held" is uninformative when there's no AIO pressure to hold against.

How we compute it

  1. Aggregate every URL in the current and prior equal-length windows from gsc_page_day.
  2. Inner-join on URL; keep rows where clicks ≥ 50 in both windows (signal threshold).
  3. Keep rows where current CTR ≥ prior CTR — CTR held or improved.
  4. Sort by current clicks, descending — the biggest pages first.

How to read each row

  • Page — the URL.
  • Clicks — current-window total.
  • CTR (current) / CTR (prior) — before-and-after.
  • CTR Δ — change in percentage points; non-negative by definition.
  • Clicks Δ — absolute change in clicks vs the prior window.
  • Avg Position — your impression-weighted position.

What to do with it

  1. Open the top 5 resistant pages and study them side-by-side. Common patterns we see:
    • Definitive answers — a direct one-sentence answer high on the page that AIO can't summarise away.
    • Strong FAQ schema with question-style H2s.
    • Distinctive brand voice or POV that a synthesised AI answer can't reproduce.
    • Depth and interactivity (calculators, comparison tables) that pull the click despite the AIO answer.
    • Branded queries where the user wants your page specifically.
  2. Identify which patterns are missing from your AIO-eroded pages.
  3. Ship the change. Re-check after 2–3 weeks — pages should start migrating from the eroded list into this list.

Caveats & limits

  • "Held CTR" is a statistical pattern, not a causal claim. A page might be resistant for the same window for unrelated reasons (a viral mention, a competitor outage). Use the list as a generator of hypotheses, not a final answer.
  • The ≥50 clicks-per-window threshold means small pages won't show up. Widen the date range to catch lower-volume URLs.
  • Branded queries can inflate CTR; if your top resistant pages are all brand pages, the patterns won't transfer cleanly to non-brand SEO pages.

Related reports