Rankings Don't Predict Citations Anymore. Here's My Proof

Something strange happened on my site lately, and it took me some time to fully understand what it meant.

My Google indexing got worse. My clicks stayed flat at basically nothing. And at the same time, my best page for bringing in actual subscribers turned out to be a page Google refuses to index at all (it all came through AI citations - more on that a little later on).

That sounds like a contradiction. It is not. It is exactly what the research says should be happening right now, and almost nobody has adjusted to it yet.

Let me show you where search is going, then show you my own numbers, because they line up in a way I did not expect.

SEO, AEO, and GEO are collapsing into one thing

Right now people talk about these as three separate disciplines. SEO to rank on search engines. AEO to get picked up in answer engines. GEO to get referenced by the models.

I don’t think that split will last for long. The sharper voices in the industry are already predicting the terms will merge, and I think they are right. In twelve to eighteen months, structuring your site for AI extraction will feel as basic as making it work on mobile, which became a default browsing method by most of us. Nobody calls that "mobile optimisation" anymore. It is just building a website properly.

So what are all three actually doing? They are the same job wearing different hats. They are trying to get you found, quoted, and recommended. That job has a name already, and it is distribution.

That is the whole reason I keep writing about this. The tactics keep getting rebranded. The underlying game does not change. Own how people find you, or rent it from someone who can switch you off.

What the new game actually rewards

If the disciplines are merging, the useful question becomes what the merged thing rewards. From the research, and from watching my own site, it comes down to four things.

1. Your pages have to be healthy. Crawlable, indexed, fast, and rendering their content without JavaScript. This is the boring floor, and it is still where most sites quietly fail. You cannot be cited if the engine cannot read you.

2. You have to be an authority on something specific. This is the part that has genuinely shifted. Backlinks used to be the currency. Now the models cross-reference everything the web says about you, across Reddit, YouTube, review sites, industry publications, and your own domain. SE Ranking found domain authority is the single strongest predictor of AI citations. Being trusted in one category now beats being present in ten.

3. Fresh content matters more than it used to. Semrush found that pages not updated quarterly are three times more likely to lose their AI citations. Publishing something and letting it sit is now an active risk. The models keep reassessing who is current.

4. You have to own a niche rather than dabble in a category. The engines are picking a small number of sources per answer. Being the obvious name in one narrow thing gets you into that set. Being vaguely present in a broad thing does not.

The number that explains everything

Here is the finding that reframed this for me.

Ahrefs looked at 863,000 keywords and 4 million AI Overview URLs. In mid-2025, about 76 percent of the pages cited in AI Overviews also ranked in Google's top ten. Seven months later, that was down to 38 percent. BrightEdge, using a different method, puts the top-ten overlap at around 17 percent, and says it has sat there all year. Roughly five out of six AI citations come from pages that are not on page one.

This means rankings have stopped predicting citations. The two systems have come apart. The page Google ranks and the page AI quotes are increasingly not the same page.

While we are on numbers, a word of caution. There is a lot of inflated data going around on this topic right now, usually from people selling the fix. I keep seeing claims that ranking first now converts about five percent of clicks, down from thirty. The independent research says position one sits around 27 percent today, down from roughly 40 percent in 2022. That is a serious decline and it deserves attention, and it is not a collapse. Ahrefs measured a 58 percent CTR drop for top-ranking pages when an AI Overview is present. Seer Interactive measured 61 percent. Those are the real numbers, and they are bad enough without exaggeration.

One more from Seer, because it points at what actually works: being cited inside an AI answer delivered two to five times the click-through of not being cited, from the same position. The citation is the prize now.

Then I looked at my own numbers

Since mid-June, my indexed pages dropped about 16 percent. The bucket Google labels "crawled, currently not indexed" almost tripled, up more than 180 percent. My impressions actually rose about 16 percent over the last month, and my clicks stayed exactly flat. My average position is around 25, which is page three, where nobody has ever gone.

Now the other half. Every single new subscriber I got this month signed up from a content page. Half of them came from one page. That page is sitting in Google's "crawled, currently not indexed" pile.

Meanwhile Copilot has cited my site 679 times in three months, and on the question about the best newsletters in my space, my pages make up nearly 58 percent of everything it quotes.

So the page Google threw away is the page bringing me readers. My worst-performing asset by Google's scorecard is my best-performing one by the only measure I care about, which is whether real people end up finding me.

I am that statistic. I am the five out of six. Buried past page three, but cited anyway.

What this means for you

The takeaways are simple, and none of them are clever.

Stop reading rankings as a health check. They are now only one side of the story and don’t paint the full picture anymore.

Measure citations, not just clicks. Open Bing Webmaster's AI Performance report. It is free, it shows real citations rather than estimates, and most people have never opened it once.

Get specific enough to be the obvious answer. I did not win "newsletters." I won a narrow question about newsletters for one kind of technical reader. Narrow is what gets you into a three-source answer.

Keep your good pages current. Quarterly updates are now maintenance, not vanity.

Fix the boring floor anyway. Crawlable, indexed, clean structure, answer in the first hundred words. All the authority in the world does nothing if the engine cannot parse you.

And do not delete anything before you check both scorecards. This nearly cost me my best page.

It is early, and that is the whole opportunity

Here is why I am not discouraged by any of this.

Almost nobody has adjusted. Most people are still staring at rankings and click counts, watching them slide, and concluding they are losing. Some of them are actually winning somewhere they are not looking.

The clicks matter less every month. The citations matter more. Very few people are measuring the thing that is growing. That gap is the opening, and openings like this do not stay open. Mobile was obvious in hindsight too, and the people who moved early spent the next decade ahead.

I would rather be early and slightly wrong than late and perfectly correct.

What I'm Thinking About This Week

Whether "SEO" as a word survives the next two years. Every part of it is turning into one question: can people and their AI find you, trust you, and quote you. That is distribution, and it is the only part of this that has ever really mattered.

Reply to this email with one of:

• Have you checked whether your worst Google page might be your best AI page? Mine was.

• Which one narrow topic could you realistically own, to the point where you are the obvious answer?

• If rankings disappeared as a metric tomorrow, what would you measure instead?

P.S. The toolkit, in the order I would use it: Visibility Audit → GEO/AEO Audit Checklist → How to Get Cited by AI → SEO Strategy Playbook → Title & Meta Audit → Sovereign Idea Workflow → Business Model Canvas. All free.

P.P.S. Know someone watching their Google numbers slide and assuming the worst? Forward this. They might be winning somewhere they have not looked yet. They can subscribe at thesovereigntechnologist.com.

That’s all for this week.

See you next Thursday.