I Measured Who AI Search Cites Across My Four Blogs, and Rebuilt My AI SEO Strategy Around the Answer

Twenty-four out of a hundred. That is the whole answer.

The honest starting point for an AI SEO strategy is that independent blogs got 24 of the 100 citations I counted in Google's AI Overviews, platforms got 20, and institutions and brands got 56 — and across the same twelve questions, ChatGPT cited an independent site exactly once in eleven answers. None of my four blogs was cited by either engine, anywhere.

So the door is open, but it is narrower than the "AI search is the new SEO" posts suggest, and it is a different width depending on what you write about. Here is the measurement, the method so you can repeat it, and the four things I changed.

What I measured, and how, so you can do the same thing tonight

On 2 October 2026 I took twelve questions — one for each content area across my four sites — and asked each one twice: once in Google Search with `hl=en` and `gl=us`, and once in ChatGPT.

For Google I expanded the AI Overview fully and wrote down every domain it linked to. For ChatGPT I recorded the source links visible in the answer. Then I sorted every cited domain into three buckets:

- I — independent: a blog or a creator site, the kind of thing you or I run.
- P — platform: YouTube, Reddit, Etsy, Pinterest and the rest of the user-generated layer.
- A — authority: an institution, a government body, a university, a large publisher or a software brand.

That is it. No tools, no subscription. It takes about an hour for twelve questions, and it is the single most useful hour I have spent on this question, because it replaces an argument with a count.

Two honest limits before the numbers. First, twelve questions is a small sample, and one evening is one evening — AI Overviews are volatile and the same question can come back differently a week later. Second, my ChatGPT account is biased: it carries custom instructions demanding verified, credible sources and it runs in Hungarian. That almost certainly pushes its answers toward institutions, so the ChatGPT half of this should be read as directional, not as what a typical US user sees. Both of those caveats cut against my own conclusions, which is why they are up here rather than in a footnote.

The count

Across twelve Google AI Overviews, 100 citations:

- Authority and brand — 56
- Independent blogs and creators — 24
- Platforms and UGC — 20

And the spread between topics mattered far more than the average. Three questions had AI Overviews where independent blogs took 45% or more of the citations: toddler activities at home (4 of 7), lunch ideas for picky eaters (4 of 7), and high-protein breakfast ideas (5 of 11). Three had zero independent citations: how sinking funds work (0 of 8), how to stop overthinking (0 of 6), and how to regulate your nervous system (0 of 8).

Look at what those two groups have in common. The open questions are practical, domestic and experience-based — somebody has actually done the thing and written it down. The closed ones are money and mind: topics where Google appears to prefer institutions, which is exactly what you would expect from a system being careful about health and finance.

That is an uncomfortable finding for this blog in particular, because EW's own subject matter sits closer to the closed end than the open one.

ChatGPT is a different door, and it is narrower

Across eleven answers where I could record sources, ChatGPT cited one to four sources per answer and almost entirely institutions: Harvard's Nutrition Source, the NIH, PubMed, the APA, the NHS, moneyhelper.org.uk, the CFPB, Canva, Etsy. In eleven answers it cited exactly one independent site — momables.com, on the picky-eater question.

One answer was more instructive than the rest: on shadow work prompts for beginners, ChatGPT ran no web search at all and answered from its own knowledge. No amount of blog writing gets you into an answer that does not look anything up.

The conclusion I drew is not "give up on ChatGPT". It is that the two engines reward different things. Google's AI Overviews can be earned by being the best practical answer. ChatGPT visibility, for a site my size, comes from being the kind of page that cites the institutions it already trusts, and from the brand and entity signals that make a name recognisable — not from out-ranking Harvard.

Why none of my sites got cited, including the part that is my own fault

The obvious explanation is that my sites are young. That is true and it is not the whole truth.

When I audited ten articles across the four blogs on 11 September 2026, eight of the eight published in May and June were one-paragraph stubs — one of them a single sentence. All three published from 31 August onward, once a proper writing process was in place, were full 1,300–1,800-word articles. Roughly 200 of about 296 URLs across the four sites predate that process.

So the site an AI engine would have found, on most URLs, was a stub. I do not think I can reasonably be annoyed about not being cited for a page that is fifty words long. That is the least flattering number in this article and it is the most useful one, because it is the only one I can act on this week.

The four changes I made

1. I stopped targeting head questions in the closed categories. If a question's AI Overview cites only institutions, writing the ninth decent article about it is spending effort where the door is shut. On those topics I go long-tail and specific — the version of the question a person actually types into an assistant, not the two-word category.

2. On the open categories, I write to be the citable answer. The direct answer in the first two sentences, numbered steps, and an FAQ written as full natural-language questions. Not because it is a clever trick, but because that is the shape the engines lift text out of.

3. I cite the institutions the engines already cite. If an AI Overview on a nutrition question is leaning on PubMed and EFSA, then a page that quotes PubMed and EFSA — accurately, with links — is a better candidate than one that quotes nobody. This costs an extra twenty minutes of fetching sources per article and is the cheapest change on this list.

4. I fix the stubs before I write anything new. A thin page is worse than no page: it spends the click, and it teaches both the reader and the crawler that the site is empty. Deleting or expanding 200 stubs is dull work with no publishing dopamine attached, and it is now ahead of new articles in the queue.

Where this sits on the chain, and the part I cannot yet measure

Everything on this blog has to sit somewhere on one chain: content → traffic → email list → a simple digital product → repeatable revenue. This measurement sits at the second link, discovery. It tells me where attention can be won.

What it does not tell me is whether any of it converts, and I want to be straight about why: there is no analytics on any of my four sites. Every tracking parameter I have appended to every link since April 2026 has gone unrecorded. So I can tell you precisely who AI search cites, and I genuinely cannot tell you how many people arrived here from where, or what they did next.

That is the gap between a measurement and a result, and I am not going to dress one as the other. The next thing I fix is the analytics, because until it exists every conclusion past the first link of the chain is a guess — including mine.

Frequently asked questions

What is an AI SEO strategy, in practice?
It is deciding which questions you can realistically be the cited answer to, and then writing pages shaped so an answer engine can lift them. The measurement above is the first half of that: before optimising anything, count who actually gets cited on your topics. Twelve questions, one hour, no tools.

How do I find out whether AI search cites my site?
Search your main questions in Google with an AI Overview showing, expand it, and read the linked domains. Then ask the same questions in ChatGPT and look at its source links. Do it for ten or twelve of your questions and write down every domain. If your site is not there, you now know where it is not there — which is more actionable than a general worry.

Does getting cited in an AI Overview bring traffic?
I cannot tell you from my own data, because my sites have no analytics installed — that is the honest answer and it is the reason analytics is the next thing I fix. Treat any confident number you read on this elsewhere with the same suspicion you should treat mine.

Is it still worth writing blog posts if AI answers the question?
On this evidence, yes on practical topics and much less so on money and mind topics, where institutions took every citation in three of the questions I checked. The decision is per topic rather than global, which is why counting beats arguing about it.

What about how to rank on Google the normal way — has that changed?
The ordinary organic results sat underneath every AI Overview I looked at, and YouTube appeared in 11 of the 12 Overviews. Nothing in this measurement suggests classic search has stopped mattering; it suggests there is now a second surface above it with different tastes.

How often should I repeat the measurement?
Monthly is plenty, and it is worth keeping the same twelve questions so the numbers are comparable. Mine are now a baseline with a date on them, which means next month's run can show a change rather than an impression.

My site is new. Is there any point?
That is roughly my position, and the stub finding above is the reason I would start with the pages you already have rather than with new ones. A hundred thin pages is a worse asset than ten good ones, for readers and for crawlers.

Get the free guide on visibility and reach

Let's Crack the Code: Hashtags & Algorithms covers the other half of discovery — how reach is actually distributed on the platforms, and what to do about it. The whole thing, free. No upsell at the end →

Get Let's Crack the Code: Hashtags & Algorithms https://ellawealthcommunity.com/lets-crack-the-code-mastering-hashtags-and-algorithms

If you want the method written down in order

Content Strategy 101 is the step-by-step version: how to decide what to publish, in what order, and how it connects to the list and the product. Seven euros. The method, written down, in order →

Get Content Strategy 101 — €7 https://ellawealthcommunity.com/content-strategy-101

Subscribe to our YouTube channel: (3) Ella Wealth Community - YouTube

Ella Wealth Community

Follow us

Contact

Subscribe to our Newsletter

ellawealthcommunity@gmail.com

Copyright © 2026 Ella Wealth Community. All rights reserved.

Privacy Policy Terms&Conditions Disclaimer Refund Policy