All articles
Local SEOAI OverviewsThin Content

Is Your Page Too Thin to Survive AI Search?

Mehr Farooq·2026-10-31· 7 min read
A hand with a red pen circling the empty middle of a printed page draft on a wooden desk, with a mug of tea, reading glasses and a small stack of papers

What made me go back and re-check every page I own?

Yesterday I published what AI Overviews actually did to my local pages, and the pattern was a split: the pages that faded were my thinnest ones, and the pages that held were the ones answering one specific local question completely. Saying that is easy. Proving it on your own site is slower, because first you have to answer an uncomfortable question honestly: which of my pages are actually thin?

So I sat down and read my own pages the way a stranger would — no memory of writing them, no mercy. It was not a comfortable afternoon. Some pages I was proud of turned out to be a heading, a few general paragraphs, and a contact button. This post is the check I ran, written down so you can run it on your own site before AI search runs it for you.

What does a too-thin page actually look like?

Thin is not about word count. I have seen 400-word pages that answer their question completely and 1,500-word pages that say almost nothing. A page is too thin when it names a question and then only half-answers it: it talks around the topic, in general terms, and stops exactly where the reader's real question starts.

The signs are consistent. There are no specifics — no real numbers, no named places, no steps, no examples. Nothing on the page could only be true of one business in one town. If you could lift the whole page, swap in a different city and a different company name, and it would still read correctly, that page is thin. It was written to contain a keyword, and it shows.

Why does a thin page lose to an AI Overview?

Because an AI Overview is, at its core, a summary of the average answer — and a thin page is the average answer. If your page says the same general things everyone else says, in roughly the same order, the summary can replace it without losing anything. The reader gets the gist from Google and never clicks. The page did not get punished. It got made redundant.

I know where that road ends because I own the number. In early October, my Houston roofing site sat at an average position of about 58.9 for "roofing contractor houston" — page six. That broad, city-wide page is the thinnest kind of page there is: written for everyone in a huge city, answering nobody's actual question. AI search did not create that problem. It just removed the last reason to pretend it was not one.

Proof — live site
epicroofingtx.com
Visit

How do I check a page before I trust it?

My check has five steps, and none of them need a tool. First, I write down the one question the page exists to answer — if I cannot state it in one sentence, the page does not know either. Second, I read only the page and ask whether that question is answered completely, or whether I would still have to search again. Third, I hunt for specifics: figures, steps, names, prices, places — anything that could not be guessed.

Fourth, I do the swap test: could this page belong to any company in any city? If yes, it fails. Fifth, I run the test I mentioned yesterday: could an AI summary quote this page with confidence? A page passes when it contains an answer specific enough to be worth quoting. Most of my older pages failed at step three. There was simply nothing on them that was not already on a hundred other pages.

The whole check takes maybe fifteen minutes a page, and most of it is just reading. That is deliberate. A tool will happily score a thin page green as long as the keywords are present — but the failure I am looking for is not a missing keyword, it is a missing answer, and only reading finds that.

An open notebook with a handwritten checklist, most lines ticked, beside a printed page and a mug of tea on a wooden table

Did other SEOs find the same thing in their data?

This is the part that turned my suspicion into something firmer. When I shared my findings on Threads, other SEOs replied with their own Search Console data, unprompted — and it matched. The pages that lost traffic were thin, top-of-funnel pages that AI Overviews absorb. The pages answering specific local queries — their example was the kind of narrow, place-and-price-specific question a real customer types — held their positions or climbed.

One of them summarised it as: AI killed the shortcut of ranking garbage. I keep coming back to that line, because it separates two very different problems. If your traffic fell because your pages were thin, that is fixable, page by page. Nobody in that conversation described careful, specific pages collapsing. The split held across different people, different niches and different data — which is exactly why I trust it more than any single study, including my own.

What do I add when a page fails the check?

I do not start by adding length. I start by completing the answer. If the page promises a question, the full answer goes on the page: the steps in order, the ranges with their sources and dates, the exceptions a real reader will hit. On my UAE salary site, the pages built that way — one profession, one place, one exact question, answered with real published figures — are the ones that were moving the right way: 52 impressions at an average position around 37 on 26 October. Small numbers, honestly reported, but earned by specific pages, not generic ones. That is the standard I now hold every page to before it ships: sourced, dated, and finished.

Then I add the local layer a summary cannot invent: the neighbourhood, the local rule, the detail you only know if the page was written for that place and not for a keyword list. And I cut whatever was only there to fill space. A fixed page is usually not longer than the thin one. It is just finished — it leaves the reader with no reason to search again, and the summary with nothing to replace.

A hand writing notes in the margin of a printed page draft on a desk, next to a folded paper street map, sticky notes and an open notebook

So how thin is too thin for your pages?

Here is the short version of the whole check. A page is too thin when it half-answers a general question with general words. A page survives when it completely answers one specific question, for one specific place, with specifics only it has. AI Overviews did not raise some impossible new bar — they removed the floor that used to hold up pages that never finished the job.

Run the five steps on your own pages before you rewrite anything, and be as unsparing with yours as I had to be with mine. Then tell me what you found — so: when you read your most important page like a stranger, which step does it fail first?

Want SEO built on proof, not promises?

I build and rank websites for home service businesses — and I test everything on my own sites first.