Ranking #1 Doesn't Mean What It Used To

For twenty years, SEO had one scoreboard. Where do you rank? Position one was the prize, position three was decent, page two was nowhere. Everyone understood it, including clients, which is part of why it survived so long as the industry’s headline number.

That scoreboard is now misleading, and 2026 is the year it stopped being a rounding error.

What changed

AI Overviews, the generated answer that sits above the traditional results, now appear in roughly a quarter of all searches. When one appears, the results underneath it move down the page and lose clicks. Measurements of how much vary considerably by study and by query type, and it is worth looking at the spread of published figures rather than trusting any single number. The direction, though, is consistent and the size is serious: top-ranked pages lose a substantial share of the clicks they used to get, with estimates commonly landing around a third and some reporting steeper drops.

The distribution is not even, which is the useful part. Informational queries, the “how do I”, “what is”, “why does” questions, trigger AI Overviews far more often than navigational ones. If someone searches your brand name, they will probably still get you as a link. If someone searches the problem you solve, they will increasingly get an answer instead of a list.

That is the shift in one sentence: you can hold position one and still lose the visit.

The part that is genuinely good news

Here is what gets lost in the doom coverage. Being cited inside an AI Overview is worth more than being listed below it.

When your page is named as a source in the generated answer, click-through rates go up, in some measurements considerably. It makes intuitive sense. A link in a list of ten is a candidate. A citation inside an answer reads as an endorsement, the source the machine chose to trust. Users clicking it are further along, more confident, and often closer to buying.

So the picture is not “AI is taking your traffic.” It is that traffic is being redistributed from everyone who ranks toward the few who get cited. That is a harder game with a better prize.

What this means for how you measure

If you are still judging your search performance solely on average position, you are reading an instrument that no longer measures the thing you care about. A few adjustments:

Stop treating position as the outcome. It is an input now. Ranking well still matters, because AI Overviews draw their sources heavily from pages that rank, but it is the beginning of the funnel rather than the end of it.

Watch clicks and CTR, not just rankings. A page holding position three while its clicks quietly halve is telling you something a rank tracker will never show. This is exactly the pattern to look for in Search Console.

Check whether you are being cited at all. Google now reports on your presence in AI experiences inside Search Console, which we covered in our piece on the new AI controls. It is impressions-only for now, but appearing at all is a signal worth knowing.

Separate your query types. Your informational content and your commercial pages are now living in different worlds. Blend them into one average and you will see a muddle instead of the actual story.

How pages actually earn citations

There is a lot of mystique being sold here. The honest version is less exciting than the marketing suggests, and closer to good writing than to a trick.

Answer the question directly, early, and in plain terms. Generated answers are assembled from passages that cleanly resolve a question. A page that spends four paragraphs warming up before saying anything useful is hard to quote. Put the answer near the top and elaborate afterwards.

Be specific. Numbers, dates, named steps, and concrete conditions get extracted. Vague, hedged prose does not. “Most businesses see results in three to six months” is quotable. “Results vary depending on many factors” is not.

Keep it current and correct. Stale figures make a page a risky source. Anything with a year in it, pricing, statistics, or platform behavior needs revisiting, not just publishing.

Structure it for extraction. Clear headings that match real questions, short self-contained paragraphs, and lists where lists make sense. Add schema markup so there is no ambiguity about what the page is and who wrote it.

Be a credible source in the first place. Attribution, author identity, and consistency across the web all feed into whether you get named. This is the slow part, and it is the part competitors cannot shortcut.

You will notice none of that is a hack. It is the same discipline that has always produced good SEO, applied with the awareness that a machine is now reading your page and deciding whether to quote you. That is why we treat classic search and AI search as one program rather than two, and why our work as a Texas SEO company looks much the same whether the reader is a person or a model.

The honest caveat

Nobody has this fully solved, and you should be skeptical of anyone claiming otherwise. The measurement is immature, click data inside AI surfaces is still limited, and the platforms are changing what they show month to month. Google was adding new citation displays and generated imagery to Overviews as recently as this July.

What is not in doubt is the direction. Answers are moving above links, citations are becoming the currency, and the businesses that adapt their content now will be the ones being quoted when the measurement catches up.

Position one is still worth having. It is just no longer the finish line.