The chart that brings people to us looks the same every time. Impressions flat or up. Clicks down, sometimes sharply. Rankings unchanged. Nothing was penalised, nothing was hacked, and the agency has no explanation beyond "algorithm update".
Nothing went wrong. The result page changed. A growing share of searches are now answered on the page itself, by a generated summary that reads several sources and cites a few of them. The user gets the answer and does not need the click. Meanwhile a second search surface has appeared entirely, inside assistants, where there is no result page at all.
This is a real change in a channel that many businesses depend on, and most of the advice about it is either panic or vendor copy. Here is what we actually see working.
The short version
- Position one is no longer the prize. Being cited is. The summary picks a handful of sources; everything below it competes for a much smaller pool of clicks.
- The clicks you keep are better. Informational queries are being absorbed; queries with intent behind them still click, because people making a decision want to see who they would be dealing with.
- The inputs did not change as much as the outcomes did. Clear structure, genuine expertise, a fast site and real citations still decide who gets used. Most of that is ordinary good SEO done properly.
- Write to be quoted, not to be scrolled. A direct answer in the first hundred words, then the depth underneath.
- Change the scoreboard. Reporting sessions alone will make a channel look like it is dying while it is quietly getting more efficient.
Why did clicks fall while rankings held?
Because a ranking is now a claim on a smaller pool. When the answer sits above the results, the queries that lose their clicks first are the ones where a sentence is sufficient: what is X, how do I spell Y, what time does Z open.
That is not evenly distributed. Query types split roughly like this.
| Query type | What happens now | What to do |
|---|---|---|
| Definitional, "what is X" | Mostly absorbed into the answer | Compete to be the cited source, not for the click |
| Comparison, "X vs Y" | Summarised, but people still click through to check | Keep these. Publish honest comparisons including where you lose |
| Local, "X near me" | Map and profile dominate above everything else | The profile is the asset, not the page |
| Commercial, "X pricing", "hire X" | Still clicks, still converts | These are the pages to invest in |
| Long-tail specific problems | Answer engines lean heavily on real detail | Your best opportunity, and the least contested |
The strategic reading: content that exists only to catch definitional traffic has lost most of its economic purpose. Content that demonstrates you can solve a specific expensive problem has gained.
What actually influences whether you get cited?
We have no privileged access to how any of these systems rank, and neither does anyone selling you a course about it. What we have is a consistent pattern across client sites, and it is unglamorous.
A direct answer, early. Systems that summarise prefer sources that state a position plainly in the first paragraphs. If your answer arrives in paragraph nine after a preamble about how the industry has evolved, you will lose to the page that answered in the first line. This is also simply better writing.
Structure that can be parsed. One H1. Descriptive H2s and H3s, ideally shaped as the question a person would ask. Short paragraphs. Lists and tables where a list or a table is genuinely the right form. Valid structured data on the page types that support it.
Specificity a general model cannot generate. Named constraints, real sequences, the exception that only shows up in practice. Generic content is exactly what the summariser can already produce without you, so it has no reason to cite it.
Citable evidence. A claim attributed to a named primary source is more quotable than the same claim floating free. Linking out to the regulator, the standard or the dataset makes a page more likely to be used, not less.
Being fast and reachable. If your page is slow, blocked, or renders its content only after JavaScript, you are optional. We costed this out in what website speed is actually costing you and the conclusion has only got sharper.
Corroboration elsewhere. These systems cross-check. A business that appears consistently across directories, profiles, reviews and third-party mentions is easier to trust than one that exists only on its own domain.
What to stop doing
- Publishing volume for its own sake. Thin definitional pages now compete directly with a machine that writes them instantly and for free. That is an unwinnable position.
- Withholding the answer to force a click. The tactic of teasing and gating now costs you the citation as well as the reader.
- Chasing keyword volume without intent. High-volume informational terms are precisely the ones being absorbed.
- Buying "AI search optimisation" as a separate product. Mostly it is ordinary technical SEO and content quality, repackaged at a higher price. Ask any vendor to name the specific thing they would change that a competent SEO would not.
How should you measure a channel that does not always end in a click?
This is where most reporting is now actively misleading. Replace it.
- Impressions and share of voice, not just sessions. Appearing in an answer is still exposure even when it does not click.
- Branded search volume over time. The cleanest available signal that people are seeing you in answers, because the pattern is: read the summary, notice the name, search the name.
- Conversion rate of organic traffic. This usually goes up as informational traffic drains away. If sessions fall and conversions hold, the channel improved.
- Assisted conversions and direct traffic. A meaningful share of people who saw you in an answer arrive later by typing the name.
- Spot checks. Once a month, run your top twenty commercial queries through the major assistants and record whether you appear and what is said about you. Crude, manual, and more useful than any dashboard currently on the market.
Judging this channel on sessions alone is the same error as judging a business on revenue while ignoring margin, which we wrote about in vanity metrics versus the numbers that predict health.
What we would do in the next quarter
- Audit what your informational content is actually earning. Some of it was already worthless and is now visibly so. Consolidate or retire it.
- Rewrite your top commercial pages to answer first. The question in the heading, the answer in the first hundred words, the depth below.
- Fix the technical floor. Speed, crawlability, server-rendered content, structured data on the page types that support it.
- Publish the three things only you know. The specifics from your actual work. This is the content with any defensibility left.
- Rebuild the report around impressions, branded search, conversion rate and manual answer checks.
- Own the surfaces you control. Profiles, directories and reviews carry more weight when a machine is corroborating you. For the local version of this, see local SEO and Google Business Profile.
Frequently asked questions
Is SEO dead?
No, but the traffic mix changed and the reporting has to change with it. The work that wins is now closer to genuine publishing than to keyword engineering, which is roughly where it was always heading.
Should we block AI crawlers?
Only if you sell content access directly. For most businesses, blocking means opting out of being the answer while competitors become it. Publishers with a subscription model face a genuinely different calculation.
Does a blog still make sense for a services business?
Yes, if it demonstrates competence on specific expensive problems. No, if it exists to catch definitional searches. The second kind stopped paying for itself.
How long before we see an effect from this work?
Technical fixes show up in weeks. Content changes take a quarter or more, because these systems need to see the page, use it, and see it hold up. Anyone promising faster is guessing.
Where to go next
If you want this run as a programme with the reporting rebuilt rather than a monthly ranking screenshot, that is SEO and digital marketing as we do it. If the technical floor is the problem, website development is where that gets fixed, and if you need pipeline while the organic work matures, PPC management covers the gap. Otherwise, talk to us.
