Is AI killing website traffic?

Short answer

For some kinds of query, yes, clicks are falling. For others, barely at all. Simple informational answers are being absorbed into AI Overviews. Local, commercial and transactional searches still send people to websites, because the searcher has to call, book or buy somewhere. Check which kind of traffic you lost before blaming AI.

This has become the default explanation for every traffic drop, and that is a problem, because it is sometimes true and often not. When it is not, blaming AI means the real cause goes unfixed.

What the data says

The clearest public numbers come from SparkToro, which found that 68.01% of US Google searches ended without a click between January and April 2026, up from 60.45% in 2024. Their analysis also put AI Overviews on more than a fifth of searches, with click-through cut sharply where they appear.

Google’s position is different. It says that when people click from results pages with AI Overviews, “these clicks are higher quality”. It does not publish data to support that, so treat it as a claim rather than a finding. Both things can be true at once: fewer clicks, and the ones that remain more likely to convert.

Who is losing, and who is not

Losing most: sites built on answering simple factual questions. Definitions, conversions, “what is” and “how many” content. If the answer fits in two sentences, it is being given in two sentences on the results page.
Losing some: broad informational content and how-to guides. AI answers summarise them, but people still click through for detail, examples and trust.
Losing little: local businesses, services and stores. Someone searching for a roofer, a lawyer or a product still has to go somewhere to act. The map pack and the product listing are still the destination.

Most of my clients are local service businesses and stores. For them, AI answers have so far been a smaller factor than the things they were already getting wrong.

How to tell if AI is behind your drop

Look at impressions and clicks separately in Search Console. If impressions are steady and clicks are falling, you are still ranking and fewer people are clicking. That pattern is consistent with AI answers, or with any other result feature pushing you down the page.

If impressions fell too, you lost rankings. That is a different problem, and AI answers are probably not the cause.

Then group the queries. If the losses are concentrated in short informational queries and your commercial pages are steady, AI is a reasonable suspect. If commercial pages fell, look elsewhere first.

Google also reports AI feature performance directly. Its guide to generative AI search points site owners to the Generative AI performance report in Search Console, which is a better source than any third-party “AI visibility” score.

What actually helps

Google is clear that there is no separate trick for AI. Its own guidance says its AI features are rooted in core Search ranking and quality systems, and that optimising for them is still SEO. Files like llms.txt are ignored by Google Search. Rewriting content in a special style for AI does nothing.

What helps is content that cannot be summarised away. A first-hand finding, a real number from your own work, a judgement call that depends on experience. A model can paraphrase a definition. It cannot invent the result of an audit you actually ran. That is the standard I hold my own writing to.

For businesses that sell locally, the bigger lever is still the basics: a properly built Business Profile, service pages that answer what buyers actually ask, and reviews that are recent.

What I would not do

Do not abandon informational content entirely. It still builds the topical depth that helps your commercial pages rank, and it is where people meet you before they are ready to buy.

Do not pay for AI visibility tracking before you have looked at your own Search Console data. The free data answers most of the question.

And do not let AI become the explanation that stops the investigation. I regularly find redirect errors, indexing problems and lost pages behind drops that had already been put down to AI.

A worked example of the check

Say your blog lost forty per cent of its clicks over six months. Export the queries for the last three months and the same three months a year earlier. Sort by clicks lost. If the top fifty losses are questions like “what is” and “how long does”, and average position on them has barely moved, the clicks went to an answer on the results page. If the top losses include your service terms and your position on them fell from four to eleven, something else happened, and that something else is usually fixable.

It takes about an hour to do this properly. It is the hour that decides whether you spend the next quarter rewriting content for a problem you do not have.

Where this is heading

I do not think anyone can tell you confidently what search traffic looks like in two years, and I would be suspicious of anyone who claims to. What I can say from the sites I work on is that the businesses holding up best are the ones that were never relying on easy informational traffic in the first place. Their search traffic was always people trying to hire, book or buy, and those people still need a website at the end of the search.

If you have lost traffic and are not sure whether it is AI or something you can fix, my SEO recovery diagnosis separates the two before anyone spends money on the wrong answer.