Our Traffic Dropped, Conversions Rose, and AI Overviews Explained Why

We manage SEO across a set of client sites, so we watch their search traffic closely. On several of them, the numbers stopped making sense. Organic traffic was sliding while conversions climbed. Same pages, same period, opposite directions.

Traffic and conversions usually move together. When they split, something is going on that your standard reports do not explain. So we went looking. ## We ruled out rankings, tracking, and intent Were rankings dropping? No. Tracking broken? No. A technical problem on the site, a shift in search intent? No and no. Every standard explanation came back clean, which is its own kind of frustrating.

One thing kept nagging at me. The Google results page itself keeps changing. It is not a list of ten blue links anymore. Depending on the query, a searcher hits ads, a local pack, shopping results, a featured snippet, video, People Also Ask, and an ## AI Overview sitting on top of all of it.

My read was that AI Overviews were the biggest factor behind the lost visitors. A page can rank exactly where it always did and still bleed clicks, because the result wrapped around it now answers the question outright. The searchers who leave, often the ones earliest in the funnel, either get what they need from the overview or click straight through to one of its cited sources. Yours, if you are lucky. Usually not. That also explained the half that had us puzzled. If the people peeling off were mostly browsers, early-stage and unlikely to buy, then the visitors still clicking through skewed later-stage and higher-intent. Fewer sessions, but a more committed slice of them. Traffic down, conversion rate up, one cause underneath both. ## The real question: what kind of result are we ranking in? “Where do we rank?” had stopped being enough on its own. What now decides whether a number-one ranking is worth anything is the kind of result it sits inside. An AI Overview can sit above your listing and take the click before anyone scrolls to you. ## Plenty of AI search advice, no keyword-level data There is no shortage of advice on optimizing for AI search. Some call it generative engine optimization, some answer engine optimization, some just AI optimization. We are still waiting to find out what the term ends up being. The advice itself is mostly fine, and mostly identical: better content, clearer answers, more authority, better structure, FAQs, schema, earned mentions.

All of it starts with tactics before anyone has looked at the situation a specific site is actually in. The industry stats had the same blind spot. Claims about how often AI Overviews show up across Google made for good headlines, but an average across the whole web cannot tell you whether your client’s best keywords trigger an overview, whether your site gets cited when they do, or who gets cited in your place.

The existing tools did not answer it either. They report inclusion, or visibility, and a visibility score with nothing under it is just a number. Inclusion only means something next to the query, the URL that actually got cited, and the competitors who beat you to it. Strip that away and you are guessing. ## So we built the AI Overview tracker ourselves For a while we did it the slow way, running searches by hand and reading the overviews one at a time. That does not scale past a handful of keywords. The data we wanted did not exist in one place, so we built the tool to collect it. The first version did one job: flag which of our tracked keywords generated an AI Overview, and whether our target site was cited. That alone earned its keep.

We built it for our own work. Then it became obvious that everyone doing this had hit the same wall, and powrank.ai was born. ##What PowRank tracks for each keyword The more data we pulled, the more the point of the tool changed. Whether a site was included turned out to be the starting line, not the finish. Inclusion is tied to everything else on the results page, and the pattern behind it is where the actual answers live. So the tool grew to show the whole picture. Today, for each keyword you track, it shows: - Whether an AI Overview appears - Every URL cited in that overview - Whether your target domain is one of them - Your organic position when you rank on the first page - The People Also Ask questions tied to the query, and the sources behind them

All of it exports for deeper analysis. Turning that data into actual content decisions is a longer story, and its own post.

## Why we started with Google AI Overviews We started with Google AI Overviews for one concrete reason: you can measure them. The other LLMs hand you nothing usable, so judging where you stand inside them is guesswork dressed up as strategy. Google AI Overviews you can see, track, and put side by side with the organic SERP. That made it the one surface where measuring first was even an option, which is the whole point. ## Measure first, then optimize AI Overviews do not retire traditional SEO. Rankings, technical health, content quality, internal linking, authority: all still load-bearing. AI search is not a replacement for SEO. It is another layer on top of it.

The reflex is to skip straight to tactics. Add FAQs. Publish more. Rewrite the page. Bolt on schema. Build authority. Every one of those can work, and every one works better when the data has told you which the results are actually asking for. Measure what is happening, build a plan from it, then optimize against the plan. In that order. ## Try PowRank on your own keywords Want to see what AI Overviews are doing to your keywords? Request access to our [Beta Testing program](https://powrank.ai/plans).