The crocodile effect in Search Console: what the gap really means

What is the crocodile effect in SEO?

The crocodile effect is the widening gap between impressions and clicks in Google Search Console. Plot both lines on the same chart and they can look like an open mouth: the impression line rises while the click line stays flat or falls. It’s a useful name for something you should investigate, but it isn’t a diagnosis.

There are three explanations worth checking. A result on the search page may answer the question before anyone clicks. A change in how impressions are recorded may make periods difficult to compare. Or your site may be appearing for more queries at positions where clicks are unlikely. The lines alone cannot tell you which is happening.

Can an AI Overview take clicks without changing your ranking?

Yes. A page may rank as well as before, or better, even as its clicks fall. If an AI Overview answers the searcher’s question above the organic results, a better title and description may do little to bring those clicks back. Nobody outranks an AI Overview in the ordinary list of results.

This matters particularly for sites with plenty of specific, niche answers. I’ve seen the gap widen on my own site and on a client site, though it is more pronounced on some sites than others. It’s a mistake to assume that every fall in click-through rate calls for a rewrite of your search snippets.

The wider evidence points in the same direction, with limits. In a study of US Google searches, Pew Research Center found that people clicked a traditional result on 8% of visits with an AI summary, compared with 15% of visits without one. That comparison does not prove what happened to a page. It does show why a stable ranking is no longer a reliable promise of stable traffic. I look at the broader shift in where search clicks are going.

For a query that Google now answers on the results page, I wouldn’t plan around winning back every lost click. I’d check whether the page can become a cited source, then look for related questions that still give people a reason to visit.

Did Search Console measurement change in September 2025?

Yes, and it can confuse this analysis. Around September 2025, Google stopped supporting the num=100 parameter, which let tools request 100 search results on one page. Automated rank checks could previously generate impressions for results deep in that list, even when a person was unlikely to reach them.

Locomotive’s account of the change describes falling impressions and an improving average position after those deep results stopped contributing as often. I saw sites where average position improved while impressions and clicks declined. That combination is a reason to inspect the dates, not to celebrate a ranking gain.

This change can make the jaws appear to close as recorded impressions fall. Its relevance to the crocodile effect is the baseline: if you compare a chart across September 2025 without accounting for it, you may mistake a reporting change for a change in demand or visibility. A later widening gap needs its own explanation.

The disruption also changed how I work. I no longer pay for high-end enterprise rank monitoring tools. I use data from several sources and build the tools I need around it. Whatever tool you use, its chart still needs a human explanation.

When are rising impressions just deep rankings?

A site can gain impressions by appearing for more long, specific queries without getting close to the results people click. Search Console can record an impression at a position that has little realistic chance of bringing a visit. More of those impressions can widen the gap even when no AI Overview is involved.

I wouldn’t draw a lesson about lost clicks from a query where a page averages position 74. At that depth, zero clicks tell us very little. Impressions at positions 60 to 99 are usually noise for a traffic decision. They can make the total look more impressive while hiding what happens on the first few pages, where a change may matter.

Average position also needs care. It combines the positions of many different queries and impressions. A change in the mix can move that average even if rankings stay steady for the searches that bring customers.

How can you tell which cause you have?

Start in Search Console’s Search results performance report. Give yourself a few minutes to separate the useful queries from the totals.

1. Find the date when the gap changed. If the shift sits around September 2025, compare the period before and after Google’s num=100 change. If it began later, investigate that later period on its own.

2. Separate branded searches from other searches. Then inspect the pages and queries that lost clicks. A sitewide chart can hide a stable brand while a group of informational pages loses traffic.

3. Compare clicks, impressions and average position for those queries. Export the query data if you need to sort it by position. Set aside deep rankings before judging whether useful visibility has grown.

4. Look at the results for the queries that matter. Check the country and device relevant to your audience. See whether an AI Overview or another prominent result answers the question, and whether your page appears as a source. Search results vary over time, so treat one check as evidence rather than a permanent label.

Google now provides a separate Generative AI performance report in Search Console. It gives you a dedicated view of your impressions in features such as AI Overviews and AI Mode. Use it alongside the ordinary performance report, which includes that visibility in its overall figures. An appearance is useful to record, but it is not the same thing as a visit.

What should you do once you know the cause?

If the gap comes from deep-position impressions, focus on the queries where your page could realistically earn clicks. Improve the page where it answers the right question but ranks poorly. Don’t spend time trying to repair a click-through rate built on position-74 impressions.

If the apparent change follows the September 2025 measurement shift, mark that break in your reporting. Judge current clicks, useful rankings and enquiries on their own terms. A reporting change does not need a content fix.

If AI Overviews are answering your important queries, check whether your content offers a clear, well-supported answer that Google can cite. Also look for relevant queries where people still need to compare, decide or act beyond a short summary. I’ve recently seen AI Overviews on betting and casino searches where I used to encounter them less often, so I would recheck the results rather than treat any query type as permanently clear of them.

I’d count a relevant citation as a visibility win, but I wouldn’t quietly equate it with traffic. The uncomfortable question is what that citation is worth to a publisher when nobody clicks. Track appearances, mentions, visits and enquiries separately. Being used as a source is not the same as being cited, and being cited is not automatically a business result. My AI visibility audit examines those signals together, so the next decision rests on more than the shape of a chart.