Google AI Mode is a conversational tab inside Google Search where Gemini answers your question directly instead of returning a list of links. It launched as an experiment in March 2025 and now serves more than a billion people a month across roughly 200 countries. For businesses, it changes the goal of search from ranking first to being quoted in the answer.
That shift is already showing up in traffic reports, and most of the advice written about it is either panic or hype. This guide is neither. Here is what AI Mode actually is, what the research says it did to clicks, how to measure it in your own Search Console, and what the evidence says actually works.
What is Google AI Mode?
Google AI Mode is a separate tab in Google Search, sitting alongside All, Images and News, that turns search into a conversation. You ask a question in plain language, Gemini answers it, and you can keep asking follow ups without starting a new search.
Three things worth knowing about where it stands today:
- Scale. It began in March 2025 as a Google Labs experiment. By Google I/O in 2026 it had passed one billion monthly users, with queries more than doubling every quarter since launch.
- Reach. It runs in around 200 countries and 98 languages, and it costs nothing to use.
- Capability. It connects to services like Canva and YouTube Music, so it can complete a task rather than just describe one.
It is not the default version of Google yet. But it is where Google ships new search features first, so treating it as a side experiment is no longer defensible.
Google AI Mode vs AI Overviews: what is the difference?
These get confused constantly, and they are not the same product. Google has confirmed they run as separate, isolated systems.
| AI Overviews | Google AI Mode | |
|---|---|---|
| Where it appears | A summary box at the top of normal results | Its own tab, separate from normal results |
| How it behaves | One answer, then the usual blue links below | A conversation you can continue |
| You still see links | Yes, underneath the box | Only a handful of cited sources |
| Who sees it | Everyone, on roughly half of queries | Anyone who opens the tab |
The practical difference matters. With AI Overviews your page can still be clicked from the results underneath. In AI Mode there are barely any results underneath, so if you are not one of the cited sources, you are effectively invisible for that question.
What AI search actually did to traffic
Here is where the research matters more than opinions. Google AI Overviews now appear on roughly half of tracked search queries. They replaced what used to be the featured snippet slot, and they created the thing everyone now calls a zero click search.
| Finding | Source |
|---|---|
| Clicks fall to 8% when an AI summary appears, against 15% without one | Pew Research Center |
| AI Overviews reduce clicks on the top result by 34.5% | Ahrefs |
| Only 1% of users click a source cited inside the AI summary | Pew Research Center |
The Pew Research Center study is worth taking seriously because of how it was built. Rather than modelling estimates, researchers analysed real browsing data from 900 US adults across 68,879 actual Google searches. Ahrefs found the same direction independently, measuring a 34.5% reduction in clicks on the top organic result.
The pattern to look for in your own data is very specific: impressions stay flat while clicks collapse. One documented case saw a B2B site lose 40% of its organic traffic in 90 days with no ranking changes, no penalty and no technical faults. Its click through rate had simply fallen from 4.2% to 2.5%. The site had not lost its rankings. It had lost its clicks.
How to actually measure it now
This is the most useful development of the year and it is easy to miss. In June 2026 Google launched Search Generative AI performance reports in Search Console, covering both Search and Discover.
Until then, measuring AI visibility meant guessing. You could see traffic falling but not whether AI features caused it. Now you get dedicated views of impressions and clicks from generative AI features, including AI Overviews and AI Mode.
If the reports have not reached your account yet, the manual diagnostic still works:
- Open the Search Console performance report.
- Compare the last 90 days against the same 90 days a year earlier.
- Export your pages sorted by change in click through rate, not change in clicks.
- Flag any page where impressions held steady or grew while click through rate fell 30% or more.
Those flagged pages are very likely losing clicks to an AI answer. One side effect worth knowing: traffic estimates in third party tools are now less reliable, because most still model clicks from ranking position, and position no longer predicts traffic the way it did.
The finding that changes the strategy
Most articles on this topic still repeat a comforting statistic from mid 2025: that around 76% of pages cited in AI Overviews already ranked in Google's top 10. The implication was reassuring. Rank well and citations follow.
That number has collapsed. Ahrefs re-ran the analysis across 863,000 keyword SERPs and 4 million AI Overview URLs in March 2026 and found only 38% of cited pages now rank in the top 10.
The reason is a mechanism called query fan-out. Instead of answering your question from the results for that question, Google breaks it into several related sub questions, searches for each one separately, then assembles an answer from what it finds. Your page can be cited for a sub question you never targeted. Equally, ranking first for the main term guarantees you nothing.
This flips the practical advice:
- Ranking is no longer sufficient. It helps, but it is now a minority of citations.
- Coverage of sub questions matters more. If your page answers only the headline question, it can be skipped for the six smaller ones the model actually searched.
- Depth beats a single optimised page. Fan-out rewards sites that cover a topic properly, not sites with one strong URL.
Does being cited actually bring traffic?
Here honesty matters more than optimism, because this is oversold everywhere.
Being cited is genuinely worth having. Brands that appear in AI answers earn more clicks per impression than uncited brands on the same queries, and branded searches that trigger an AI Overview can see click through rate rise rather than fall.
But keep the scale in perspective. Pew found only 1% of users click a source inside an AI summary. Citation buys you brand visibility, credibility and presence at the moment of decision. It does not replace the traffic volume you lost. Anyone promising that citations will restore your old click numbers is selling something.
So the honest answer to "is SEO dead" is no. The work changed shape, and the future of SEO looks more like disciplined fundamentals plus better structure than like a brand new toolset.
What SEO practitioners actually say about it
Most agency content here reads like a sales pitch. The real conversation among people doing the work is more sceptical, and worth hearing before you spend money.
The GEO backlash is real. Many working SEOs think generative engine optimization, and the answer engine optimization label alongside it, has been oversold as a brand new discipline. The most upvoted reaction in professional communities tends to be blunt: solid fundamentals still win, and a lot of the surrounding advice is marketing. A common prediction is that the industry will simply rebrand SEO as GEO rather than genuinely change what it does.
Some popular tactics do not work. Ahrefs analysed 137,210 domains and found 97% of llms.txt files receive no requests at all. Slack's link preview fetcher hit those files more often than Perplexity's search bot did. If someone is selling you an llms.txt file as AI optimisation, that is close to a waste of money.
There is real user backlash too. DuckDuckGo reported installs up 30% as people actively looked for a way to opt out of AI results. And a German court ruled that Google can be directly liable for false claims in its AI Overviews, on the basis that generating an answer is publishing rather than linking.
None of this means ignore AI search. It means be selective about which advice you pay for.
How to rank in AI Overviews and get cited
Cutting the noise, these are the tactics with real evidence behind them.
Cover the whole topic, not one keyword. Because of query fan-out, answer the sub questions around your subject on the same page or in a connected cluster. This is now the highest leverage change you can make.
Answer in the first 50 words under every heading. Put the direct answer immediately below the heading, then expand. AI systems extract passages, so each passage has to stand alone.
Add statistics, quotations and citations. The original Princeton GEO research, which coined the term and tested roughly 10,000 queries, found that adding cited statistics and quotations lifted visibility in AI answers by 22% to 41%. This is the best evidenced tactic in the entire field.
Use comparison tables. Structured tables get pulled into answers more readily than the same information written as prose.
Write headings as real questions. People ask AI Mode full questions, so pages organised around questions map more cleanly onto what gets asked.
Add schema markup. Structured data helps machines parse your page. FAQ schema is worth adding wherever you have genuine questions, and article and organisation markup help establish who you are.
Strengthen entity and E-E-A-T signals. Consistent naming, a clear About page, real author details and mentions on reputable sites all help systems trust you. Google has published its own guide to optimising for generative AI features, and it reads like conventional quality advice, because it largely is.
Build topical authority. Covering one subject properly and linking those pages together does more than scattering unrelated posts.
For the wider strategic picture on how classic search, answer engines and generative engines fit together, our guide to SEO vs AEO vs GEO covers the differences properly.
It is not only Google
Focusing your AI search optimization on Google alone leaves gaps, because your buyers ask other systems too.
- ChatGPT sends real referral traffic, and roughly one in five ChatGPT interactions still ends in a Google click, so the two feed each other. Earning ChatGPT citations comes down to the same fundamentals: be findable, be specific, be quotable.
- Perplexity shows its sources openly, which makes it the easiest place to test whether your pages are citable at all. Search your core topics there and see who gets named.
- Gemini shares Google's index and infrastructure, so work that improves Google visibility usually improves Gemini visibility too.
The takeaway is reassuring. There is no separate playbook per engine. Good SEO for AI search is one discipline applied consistently, then formatted so a machine can lift a clean answer out of it.
What this means for businesses in the UAE
AI Mode is live here in both English and Arabic, so this is not something arriving later. Two points matter most locally.
Bilingual signals need to be right. An Arabic and an English version competing against each other weakens both, and neither gets cited.
Local intent still behaves differently. Searches with clear local intent continue to surface maps and business listings prominently, so your Google Business Profile is still doing real work. If you need the underlying search foundations fixed, that is what our SEO services in Dubai and Abu Dhabi are built around.
The bottom line
Google AI Mode is not a preview of the future. It is a product with a billion users, and it absorbs clicks that used to reach your site.
But the response is not to chase every new acronym. Only 38% of cited pages now rank in the top 10, which means ranking alone will not carry you, and query fan-out means depth across a topic beats one perfectly optimised page. Add cited statistics, structure your answers so they can be lifted cleanly, and cover the questions around your subject rather than just the headline one.
If your traffic has fallen while impressions held steady, that is the signature, and it is worth a proper look. Talk to COM8 STUDIO and we will tell you honestly whether AI search is your problem.
