Does SEO still work with AI overviews?

I get asked does SEO still work with AI overviews at least once a week now, usually by a client who saw a traffic dip on one page and assumed the entire discipline stopped functioning. The honest answer is that SEO still works, but what working produces has changed shape, and setting the right expectation here matters more than it did two years ago.

AI overviews and AI-generated answers pull from the same underlying signal SEO has always been built on: pages Google has crawled, indexed, and judged trustworthy enough to rank or cite. Nothing about that mechanism changed. What changed is what happens after a page clears that bar. A query that used to send ten clicks to the top organic results now answers some of those searchers directly on the results page, and only a portion click through afterward.

What actually shifted, and what did not

The ranking mechanism is unchanged. Relevance, authority signals, and technical accessibility still determine whether a page is eligible to show up at all, whether inside an AI overview or in classic blue links. A page that is not indexed does not get cited by an AI overview any more than it ranks in a normal result. The eligibility criteria are the same criteria they always were.

What changed is the shape of the traffic curve once a page clears that bar. Informational queries, the kind with a single clean factual answer, are the most likely to get fully answered inside the overview with no click needed. Queries involving genuine ambiguity, comparison, or a decision about a specific business are far less likely to be satisfied by a short AI summary, because the searcher still has to visit a site to decide.

This is why a blog post answering a simple factual question can lose click volume even while ranking well, and a service page or comparison page can hold or grow its traffic over the same period. The content type determines exposure to this shift more than the SEO work behind it does.

A worked example of the split

Take a query like how much does a service cost in a given city. Before AI overviews became common on that kind of search, a page ranking well for it might have earned clicks from most of the people who searched it, because the answer required reading a specific business's pricing structure. After an AI overview started summarizing typical price ranges for that category, some of that traffic gets absorbed by the summary itself, since a searcher who just wants a ballpark figure no longer needs to click anywhere.

Now compare that to a query comparing one service against a competitor for a specific situation. That query cannot be fully answered by a general summary, because the right answer depends on specifics about the searcher's own situation that a generic AI overview has no way to know. Pages built around that kind of comparison keep earning clicks at close to the same rate they always did, because the AI overview itself, when one even shows up, tends to point toward the comparison rather than replace the need to read one.

The pattern holds across the sites I work with. Pages answering a fact lose some click share. Pages answering a decision mostly do not. Knowing which category a given piece of content falls into, before you write it, is a more useful planning tool now than it was three years ago, when almost every ranking page converted clicks at a similar rate regardless of its type.

Setting the expectation: does SEO still work with AI overviews in the room

The expectation I set with clients now has two parts, and both hold true at the same time. Total click volume on purely informational content is a less reliable growth lever than it was three years ago, because more of that demand gets satisfied without a click. At the same time, visibility itself, being the source an AI overview cites or the page that ranks when a searcher does click through, still requires the same technical and content fundamentals SEO has always required.

In practice, the content mix matters more than it used to. A site that only publishes broad informational posts is more exposed to this shift than a site that pairs those posts with service pages, comparison content, and pages built around a decision a person still has to make. That decision-stage content was already the more valuable traffic before AI overviews existed. It has simply become the more resilient traffic now.

This is also part of what I cover in a first month of any engagement, because a client who understands upfront which of their pages sit in the exposed category and which sit in the resilient one reads their own traffic data correctly from week one instead of assuming the worst at the first dip. What to expect in month one of SEO walks through that kind of baseline conversation in full, AI overviews included.

The mechanism behind AI citation

Being cited in an AI overview is not a separate skill from ranking. The same signals that get a page indexed and ranked, clear structure, direct answers stated early, accurate and specific information, are what a language model draws from when it assembles a summary. A page that is vague, padded, or written to satisfy a keyword rather than answer the question performs worse in both systems for the same reason: it does not give a clean, extractable answer to pull from.

Structurally, this rewards the same habits that have always separated a page that ranks well from one that merely exists. A clear heading that states the actual question, an answer given in the first sentence or two under that heading rather than buried after several paragraphs of preamble, and specific numbers or steps instead of vague generalities all make a passage easier for both a search algorithm and a language model to lift cleanly. Pages written to pad word count with repeated framing before getting to the point tend to perform worse in both systems, not because either one penalizes length directly, but because neither can find a clean passage to extract from a paragraph that takes four sentences to say something a single sentence could cover.

This is why I have not changed my actual technical or content approach because of AI overviews. I have changed which content types I prioritize for click-driven growth versus which ones I accept will get partially absorbed into an on-page answer. Both still need the same underlying work to be eligible in the first place.

This shift changes what to expect from a campaign, not whether it's worth running. How to know if SEO is working before rankings move covers the earlier signals worth watching regardless of AI overviews, and how much SEO should cost covers pricing that work correctly. More in the SEO Expectations archive.

When the content mix itself is the real problem

Occasionally what looks like an AI overview problem is actually a business model problem that predates AI overviews entirely. A site whose entire traffic and lead strategy depends on ranking broad informational content, with no service pages, no comparison content, and no pages built around an actual buying decision, was already carrying more risk than a site with a mixed content library. AI overviews did not create that risk. They exposed it faster.

If that describes your situation, the fix is not more informational content aimed at the same queries that are already getting summarized. It is building out the decision-stage pages that were probably missing before AI overviews ever became a factor, and when SEO won't fix your business problem is worth a look if you are unsure whether the gap you are seeing is a content mix issue or something more fundamental about how the business generates leads in the first place.

How this plays out differently by industry

The exposure to AI overviews is not uniform across industries, and I have started factoring this into content planning at the category level rather than the page level. A local service business selling something people research once and then hire someone to do, a roof repair or a plumbing fix, tends to see less overview activity, because the searcher's real need is finding a specific provider, not learning a general fact. A publisher or informational site covering broad how-to topics with no transactional intent behind them sees the most overview activity, because that is exactly the kind of query a short summary can fully answer.

B2B sites sit somewhere in between. A query about what a category of software does might get summarized, but a query comparing specific vendors for a specific use case rarely gets fully answered by an overview, because the decision depends on details a generic summary cannot account for. Understanding which bucket your core queries fall into before building a content calendar around them saves months of producing content aimed at traffic that was always going to get absorbed by a summary rather than converted into a click.

What to actually watch

If AI overviews are a real factor in your industry, do not judge SEO health by total click volume on informational content alone. Watch impressions in Search Console for those pages. A page with high impressions and declining clicks usually means its query is a candidate for an overview answer, not that the SEO work stopped functioning. Compare that pattern against your decision-stage pages, service pages, and comparison content, where impressions and clicks should still move together the way they always have.

SEO in an AI search environment still rewards the same fundamentals. What changes is which content format captures the value those fundamentals produce, and that is a content strategy question, not a sign the discipline stopped working.