Search has always been a bit odd.
For years, the deal was fairly easy to understand. Someone searched. Google showed a list of links. The business tried to rank as high as it sensibly could for the searches that mattered. People clicked, compared, read, bought, enquired, or vanished into another tab like nature intended.
That model has not disappeared.
But it has changed.
AI overviews have made the answer more visible before the click. For some searches, the user now sees a generated summary, cited sources, follow-up prompts, and enough information to delay or avoid visiting a website at all.
That is not a small cosmetic change.
It changes what visibility means.
It changes how informational searches behave.
It changes what should be monitored.
And, naturally, it has encouraged a whole new category of online shouting about SEO being dead. Again. Search has apparently died more times than a badly managed staging server.
The more useful view is less dramatic.
SEO is not dead. The old reporting shorthand is weaker. Rankings still matter, but they no longer explain enough on their own. Traffic still matters, but not all lost clicks mean lost value. Content still matters, but only if it is useful enough to be read, cited, trusted, or acted on.
The work has changed shape.
The answer increasingly appears before the click
The biggest practical shift is simple: more of the answer can appear on the results page.
That was already true before AI overviews. Featured snippets, local packs, knowledge panels, maps, shopping results, video blocks, and "people also ask" boxes had been eating the clean blue-link model for years.
AI overviews are part of that longer direction, but they make it more obvious.
For certain searches, especially broader informational ones, the search result can now do more of the explaining. The user may not need to click ten articles to build a basic understanding. They may get a summary first, then choose whether any source deserves more attention.
That affects two things at once.
First, some clicks get compressed. A visitor who once clicked three basic articles might now read the overview and move on.
Second, the sources that are cited or reinforced by the overview gain a different kind of visibility. They may not always win the same click they used to win, but they can still help shape trust, brand recognition, and later decisions.
That distinction matters.
If your reporting only asks "where did we rank?", it misses part of the picture.
If it only asks "how many clicks did this page get?", it may also miss part of the picture.
The better question is: where are we visible, useful, cited, trusted, and still converting the right people?
Less tidy. More accurate. Annoying, but that is usually how measurement improves.
Not every query is affected in the same way
This is where blanket advice gets lazy.
AI overviews do not affect every search equally.
Some searches are vulnerable because they ask for a general explanation:
- what is X?
- how does Y work?
- benefits of Z
- basic comparison queries
- early-stage research questions
Those are the searches where a generated answer can satisfy a lot of intent before the user reaches a site.
Other searches still have stronger reasons to click:
- local service searches
- product and pricing research
- high-intent commercial comparisons
- case study or proof-led queries
- queries where trust, detail, or specificity matters
- searches where the user needs a provider, not just an answer
If someone is looking for a web design agency in Leeds, a generated overview may shape the field, but the buyer still needs to evaluate actual businesses. They need proof. They need fit. They need examples. They may need to get in touch with someone who appears to have done the work rather than merely described it.
That is where SEO remains very much alive.
The top of the funnel has got noisier and less predictable. The commercial middle and bottom still matter enormously.
The trick is not to keep doing the same thing while hoping the results page behaves like it did five years ago.
The trick is to understand which searches are now mostly answer-shaped, which are still decision-shaped, and which ones carry enough commercial intent to deserve proper investment.
Rankings alone do not explain visibility
Rankings are easy to understand, which is why people like them.
Position three feels better than position nine. Page one feels better than page two. Everyone understands the little emotional injury of dropping below a competitor with a worse website and a suspiciously cheerful blog.
Rankings still matter because they are a useful signal of visibility.
They are just not the whole signal.
A ranking report can tell you where a page appears in traditional results. It may not tell you:
- whether an AI overview appeared above it
- whether your brand or page was cited
- whether another source was used to answer the question
- whether the click-through rate changed because the result page changed
- whether traffic dropped but lead quality improved
- whether users now discover you earlier but convert later
This is why SEO reporting needs a wider view.
Not a more complicated dashboard for the sake of it. Nobody needs a control panel that looks like it was designed by someone paid per widget.
Just a better set of questions.
Where do we still appear?
Where are we being cited?
Which pages still bring qualified traffic?
Which topics generate enquiries rather than vanity visits?
Which searches are now mostly research noise?
Which pages help a buyer make a decision even if they are not the first click?
That is a more useful conversation than "traffic is down, panic accordingly".
Useful pages still earn attention
It is tempting to treat AI overviews as a complete reset.
They are not.
A lot of the fundamentals still matter, partly because AI systems still need material to work from, and partly because humans still need reasons to trust you.
The basics have not suddenly become optional:
- technically sound pages
- clear site structure
- fast, accessible pages
- useful content with a real purpose
- specific answers to specific questions
- internal links that help people move sensibly
- evidence, examples, case studies, and proof
- business information that is consistent and easy to understand
If a page was thin, vague, slow, badly structured, and written mainly to satisfy an old keyword checklist, AI overviews did not create the problem. They just made the weakness harder to hide.
The work still starts with being useful.
I appreciate this is less exciting than announcing a new secret playbook. Sadly, most durable work is like that.
What makes content more likely to keep working
The pages that still earn attention tend to do more than define a term.
They help someone make progress.
That might mean:
- explaining a concept clearly enough to be cited
- comparing options without pretending every option is equally good
- showing first-hand experience
- answering the awkward follow-up questions
- making local or sector-specific relevance obvious
- backing claims with examples
- connecting advice to a commercial next step
Generic articles are easier to replace with a summary.
Specific, useful, experience-led pages are harder to flatten.
For example, a basic article on "what is business process automation?" is not very defensible. There are hundreds of those. Most are beige. A generated answer can do a passable job because the material is generic.
A page explaining which business processes are worth automating first, with examples from real operational work, common failure points, review boundaries, and buying advice, is stronger.
It gives the search engine more useful material.
It gives the reader more reason to stay.
It gives the business a better chance of being remembered.
That is the standard now.
Not more content. Better evidence.