Bing Tests AI Mode Name for Copilot Search: What to Do

26 September 2026 · The Cresia team

Probably not much, at least not yet. Microsoft is testing the name AI Mode on the tab where Bing currently offers Copilot search, as Search Engine Roundtable reported. The label may change; the way answer engines find, read and cite your pages has not.

Key takeaways

Bing is testing AI Mode as the label for its Copilot search tab. Search Engine Roundtable described it as a test, so treat it as unsettled.
Google already uses the name AI Mode. If both engines share it, any report that says AI Mode without naming the engine becomes unreliable.
Answer engines still choose sources on fundamentals: they can crawl the page, a passage answers the question cleanly, the entity is clear, and other sources agree.
The useful response is durable work: crawl access, self-contained passages, entity hygiene and a prompt set tracked per engine.
Skip the panic. A rename is not a strategy change, and a Bing-only playbook built around a label is wasted effort.

What Bing is testing with the AI Mode label

Bing has a search experience it calls Copilot Search, reachable from a tab in the results interface. According to Search Engine Roundtable, Microsoft is now testing a different name for that tab: AI Mode. That is the name Google uses for its own AI search experience.

That is the whole development. There is no reported change to how Bing retrieves pages, how it writes answers or how it chooses citations. Whether the test reaches everyone, reaches some regions, or disappears in a week is not something anyone outside Microsoft can say. Test labels in search interfaces come and go, and this one could be reversed as quickly as it appeared.

It is still worth a paragraph of your attention, because naming shapes how teams talk, and how they talk shapes how they report. When a category label gets borrowed, the confusion lands on the people who have to write the monthly deck.

Why a shared name matters for AI search reporting

Picture a head of SEO who has a slide titled AI Mode visibility. Today that slide almost certainly means Google. If Bing's tab starts saying AI Mode too, then a colleague in another region, running spot checks on Bing, might paste screenshots under the same heading. Nobody lied. The data is simply mixed.

This is the real risk, and it's a data hygiene problem, not a ranking problem. The two products are different engines with different indexes, different retrieval behaviour and different audiences. A brand can be cited often in one and rarely in the other. Averaging them hides exactly the information you need.

Search Engine Land has published several pieces recently on how AI Mode and AI Overviews affect clicks and where they appear. Those headlines are about Google's product. If your team reads them and then sees Bing's tab carrying the same name, it is easy to assume the findings transfer. They may not. Attribute each finding to the engine it studied, and keep that habit in your own internal notes.

The practical fix is small: every observation you record gets an engine field, a surface field and a date. Name the engine first, the feature second. Google AI Mode and Bing Copilot Search, or Bing AI Mode if the test sticks, are different rows.

How answer engines choose and cite sources

The mechanics matter more than the label, so it helps to be plain about them. The exact ranking and citation logic of each engine is proprietary, and it changes. What you can rely on is the general shape, which is documented in each vendor's own guidance and visible in how answers behave.

An answer engine that cites sources generally does some version of four things:

Retrieve candidates. It runs one or more searches, often broader or more varied than the user's typed query, and pulls pages from its index. If your page isn't in the index, nothing else on this list matters.
Read passages, not pages. The model works with chunks of text. A page with one strong, self-contained paragraph on a question can beat a longer page where the answer is scattered.
Weigh trust and agreement. Claims that several independent sources support are safer to state. Entities the engine recognises, with consistent names and attributes, are easier to use.
Compose and attribute. The answer is written, and some passages get linked. Which ones get linked is the part that varies most between products.

Notice that none of these steps depends on the tab's name. It also explains why Bing's index matters beyond Bing. Microsoft's search index has long been a source that other products draw on, and Microsoft's own Bing webmaster documentation is the place to check what feeds what. If a change in Bing's index or crawling affects you, it can show up in several places at once, whatever those places call themselves.

What to change in your AEO and GEO work this quarter

None of this needs a new programme. It needs a short list done properly. Here is the order we'd tackle it in for a mid-size site with a small team.

Confirm Bing can crawl and index your important pages. Many teams verify Google carefully and never look at Bing Webmaster Tools. Check that key templates are indexed, that your sitemap is submitted, and that nothing in robots rules or bot protection blocks Bing's crawler by accident.
Rewrite the top of your best pages as answers. For each priority question, put a two-to-four sentence answer directly under the heading that asks it. Then add depth below. A passage that reads correctly when lifted out of the page is the unit engines actually use.
Make entities boring and consistent. Product names, company name, founders, locations and categories should be spelled the same way on your site, your profiles and your structured data. If you rename a product, update all of them together.
Mark up what is true and visible. Structured data should describe content that is on the page. Don't add schema for claims you don't make in the copy.
Put proof where it can be found. Original numbers, named examples and dated statements give a model something specific to quote. Vague copy gets paraphrased and rarely credited.
Build a fixed prompt set. Twenty to fifty real questions your buyers ask, run on a schedule, each logged against the engine that answered.

Step six is where most of the work lives. It's also where teams cut corners, so it gets its own section.

How to measure visibility without mixing engines

AI answers are variable. Ask the same question twice and you may get different citations. Any single screenshot is an anecdote. A tracked set of prompts, repeated over weeks, is a signal.

The table below is a starting measurement plan. It assumes you're tracking at least Google and Bing surfaces, and it keeps the two apart on purpose.

Measure	What you record	Engine field	Owner	Cadence
Citation presence	Whether your domain is linked in the answer	Required, never blank	SEO lead	Weekly
Mention without link	Brand named but not cited	Required	Content lead	Weekly
Answer accuracy	Whether the answer states your product facts correctly	Required	Product marketing	Monthly
Referral sessions	Visits from AI surfaces where the referrer allows identification	By referrer source	Analytics lead	Monthly
Index coverage	Priority URLs indexed in Google and Bing separately	One row each	Technical SEO	Monthly

Two cautions. First, referral traffic from AI surfaces is patchy: some clicks arrive without a clear referrer, so treat those numbers as a floor, not a total. Second, don't compare your citation rate to a benchmark unless you know how it was measured. Reports such as the Conductor AEO/GEO benchmarks are useful for context, but their prompt sets and methods aren't yours.

If your reporting spans several sites, regions and tools, the taxonomy matters more than the tool. A tracking specification that defines what each source and medium value means, and who may add new ones, keeps a label change from quietly rewriting last quarter's chart. Our explainer on what a tracking specification is covers the basics, and teams who want to enforce that discipline across properties can look at OmniSpec.

For visibility across classic and AI search in one place, MediaPilot is the Cresia product to look at. Check the product page for what it covers rather than assuming from the name.

What to ignore in the rename

There's a temptation, whenever an engine changes a label, to draw a conclusion from it. Resist most of them.

Ignore the idea that Microsoft is conceding something. A name is a name. It tells you about interface choices, not about which engine is better or which will win.
Ignore advice to optimise for AI Mode as if it were a single thing. If two engines use the name, that advice is meaningless until it says which one.
Ignore claims of a secret ranking factor tied to the rename. Nothing reported so far suggests one.
Ignore your own urge to rebuild the dashboard this week. Wait until the test either spreads or dies.

What isn't safe to ignore is the wider trend the rename sits inside: search engines are folding conversational answers into the main interface and giving them prominent tabs. Google's AI Mode and Bing's Copilot Search both point that way. That is a reason to keep your answer-ready content in good shape. It isn't a reason to react to a label.

What not to do while the naming settles

Some mistakes are easy to make under this kind of ambiguity.

Don't fork your content by engine. Writing one version of a page for Google's AI Mode and another for Bing's tab doubles the maintenance and, as far as anyone has documented, buys nothing. Pages that are clear, sourced and crawlable serve both.

Don't stuff pages with question-and-answer blocks that repeat the same point in different words. Engines, and readers, can tell. One good answer per question beats five thin ones.

Don't publish invented numbers because a model seems to like statistics. A made-up figure that gets quoted back at you, with your domain attached, is a liability. Use your own data, name the source, and date it.

Don't treat a single screenshot as evidence, in either direction. A colleague who sees your competitor cited once in the new Bing tab has an anecdote. Log it, then look again next week.

And don't let the search team, the analytics team and the paid team each invent their own label for this. Agree the vocabulary once. A shared glossary, with engine names spelled out, is cheap. Teams that run paid and organic together can start from media teams resources to see how that joins up.

A worked example: one question, two engines

Take a B2B software company that sells scheduling software to clinics. Their priority question is: how do clinics reduce no-shows without calling every patient?

On Google's AI Mode, the answer cites a competitor's guide and a trade association page. On Bing's Copilot Search, or whatever it's called by next month, the answer cites the company's own blog and a forum thread. The team could conclude that Bing likes them. A more useful reading: their page is indexed and passage-ready in Bing, and something about the Google result is off. Maybe the page isn't ranking well enough to be retrieved, maybe the answer sits too deep in the article, or maybe the competitor simply has more corroboration.

That comparison is only possible if the two rows are kept apart. Merge them under AI Mode and the team sees one blended result that hides both the win and the gap.

The next steps are ordinary. Move the direct answer to the top of the page. Add the clinic-specific detail that only they have. Check that Google has indexed the latest version. Re-run the prompt in two weeks and compare.

Frequently asked questions

Is Bing replacing Copilot with AI Mode?

Not definitively. Search Engine Roundtable reported that Microsoft is testing the AI Mode name on the Copilot search tab in Bing. A test can be reversed or limited to a small group of users, so check what you see in your own results before changing anything.

Do I need a separate optimisation strategy for Bing's AI search?

No. The fundamentals overlap heavily: indexed pages, clear passages, consistent entities and credible supporting sources. What you do need is separate measurement, so you can see how each engine treats you. Verify Bing indexing through Bing's own webmaster tools.

Does the rename affect how my brand gets cited?

There's no reported evidence that it does. The rename concerns the label on a tab, not retrieval or citation behaviour. If citations shift, it will show up in your tracked prompts, which is why keeping a per-engine log matters.

How should I label AI search in my reports?

Name the engine first and the feature second, for example Google AI Mode or Bing Copilot Search. Record the date of each check. That keeps history readable even if a vendor renames a feature halfway through a quarter.

Should I change my content because of this news?

Only if it exposes gaps you would want to fix anyway. Check crawl access on Bing, put direct answers near the top of key pages, and keep entity details consistent. Those steps hold up under any naming outcome.

Sources

https://www.seroundtable.com/bing-ai-mode-search-tabs-42165.html
https://searchengineland.com/topic/generative-engine-optimization
https://developers.google.com/search/blog
https://www.conductor.com/academy/aeo-geo-benchmarks-report/