Two different things get called “the AI thing in Google”, and the confusion causes a lot of unnecessary worry. This article separates them, answers the question most people arrive with, and then covers what it means if you rely on organic traffic.
What is the difference between AI Mode and AI Overviews?
AI Overviews are the AI-written summaries that appear at the top of a normal results page, above the blue links. You did not ask for one; Google decides when the query warrants it.
AI Mode is a separate tab alongside Images, News and Shopping. Open it and the whole experience becomes conversational: you ask a question, get a written answer with links, and can ask follow-ups that keep the context.
The difference that matters commercially is consent. AI Overviews change the results page for everyone on the queries that trigger them, while AI Mode only affects users who deliberately choose it.
Under the bonnet they work similarly. The question gets expanded into several related queries, results are retrieved for each, and an answer is synthesised from the passages that come back. Google’s own explanation of this sits in its AI features documentation and the succeeding in AI search post.
Can you turn AI Mode off?
Most people asking this actually mean AI Overviews, so here is the direct answer to both.
AI Mode: yes, in the sense that it is a tab you choose to open. Ignore it and your normal search experience is unchanged.
AI Overviews: no. There is no setting in Google Search that switches them off.
What you can do is filter. On a results page, the Web filter returns a plain list of links with no AI summary. On desktop it usually sits under the “More” menu in the row of tabs, and you can make it stick by setting a browser search shortcut that goes straight to the Web results.
Third-party browser extensions also exist that strip AI Overviews, and browsers other than Chrome give you more control over your default search behaviour. Nothing here requires a Google account change, and none of it affects what your customers see.
That is the whole answer to the switch-off question. The rest of this article is about the business implication.
What does AI Mode change about the results page?
Three things, in descending order of impact.
The answer arrives before the links. For a question with a factual answer, the user has what they came for without leaving the page.
The links that remain are fewer and lower. Where a page of ten organic results used to compete, an AI answer cites a handful of sources and the traditional listings sit below the fold.
Follow-up questions replace new searches. In AI Mode, the second and third question stay inside the conversation, so the sequence of searches that used to produce three chances to be found now produces one.
What has it done to click-through rates?
The most useful evidence here comes from analysis of real browsing behaviour rather than from vendor surveys.
The Pew Research Center found that users clicked a search result on around 8% of visits where an AI summary was present, compared with around 15% of visits where one was not. That is roughly half.
Treat that as directional rather than as your number. The effect varies enormously by query type, and your own Search Console data is the only source that tells you what is happening to you specifically.
The pattern to look for in your own data is impressions holding steady while clicks fall. That gap is the signature of an answer being given on the results page.
Which types of query are most affected?
| Query type | Exposure | Why |
|---|---|---|
| Definitional (“what is X”) | High | The answer is short, factual and fully satisfiable on the page |
| How-to and instructional | High | Steps summarise well, and follow-ups stay in the conversation |
| Comparison (“X vs Y”) | Medium | Summarised well, but users often still want detail and evidence |
| Commercial investigation | Medium | Pricing and shortlists get summarised, and buyers still click through |
| Transactional (“buy X”) | Low | Intent is to complete an action on a site, not to read a summary |
| Local (“X near me”) | Low | Map results and profiles still dominate |
| Branded | Very low | The user has already chosen where they are going |
If your traffic is concentrated in the top two rows, you are exposed and should plan for it. If it sits in the bottom three, the effect on you is real but modest.
What does this mean if you rely on organic traffic?
Be honest about the shape of the change: it is a redistribution, not a collapse.
Some of the traffic that has gone was never commercially valuable. A visitor arriving for a one-line definition and leaving in nine seconds was a number in a report rather than a customer.
What has genuinely got harder is top-of-funnel reach. If your model was to attract people with informational content and convert them later, the first step of that funnel is narrower than it was.
What has got more valuable is being the source that gets cited, being strong on the queries that still click, and having demand that does not depend on a search results page at all.
What should you do about it?
- Segment your Search Console data by query type. Split informational, commercial, transactional, local and branded, and compare clicks and impressions over the last twelve months. This tells you your actual exposure rather than the industry average.
- Identify your most exposed pages. High impressions, falling clicks, informational intent. That is the at-risk list.
- Decide which ones are worth defending. Some informational pages exist to support a commercial page and are fine as citation sources even without clicks. Others were traffic for its own sake and can be left alone.
- Rewrite the survivors to be citable. Direct answer in the first 60 words, question-led headings, tables where you are comparing things, sources named. The full list is in our AI search optimisation checklist.
- Shift production towards queries that still convert. Comparison, pricing, local and bottom-of-funnel content is where the clicks still are.
- Start measuring citations, not just clicks. If you are the source in an answer that never gets clicked, that has value your analytics cannot see. Our method for measuring it is in how to measure AI search visibility.
One caveat on everything above. This area changes faster than any other part of search, and details like UK availability and feature behaviour shift regularly, so verify anything you are about to make a decision on against Google’s current documentation.