Microsoft Adds an AI Answers Off Switch to Bing: an Extension and the -ai Operator

For several years Microsoft has been weaving AI into Bing: first a chat assistant, then AI-generated answers right above the regular results. Now the company has unexpectedly acknowledged the obvious. According to its own research, users want to decide for themselves whether AI takes part in their search. Some want a short summary from a neural network. Others are used to opening several sources, comparing them and drawing their own conclusions.

Two Ways to Turn Off AI in Bing

The first way is the Microsoft Bing AI Search Choice extension. It is currently in preview and works in Google Chrome and Microsoft Edge. Once installed, a simple toggle appears in the browser: AI answers in Bing are either on or off. The setting is saved for all queries, so you don't have to change it every time.

The second way requires no installation: just append the "-ai" operator to your query. For example, the query "best laptops 2026 -ai" returns the classic results without AI-generated answers. It is a familiar trick that works the same way as negative keywords in regular search.

Where the Catch Is

Both methods have drawbacks, and they are not minor.

When installed, the extension makes Bing the default search engine. If you only want to control AI in Bing and leave the rest of your browser settings alone, that is an unnecessary requirement. Critics rightly note that Microsoft could have added a toggle to Bing's own settings or even directly on the results page. Then AI answers could be turned off permanently without third-party software.

The "-ai" operator is convenient for only a single query. You have to type it every time, and for everyday searching that gets old fast. The result is a formal option to choose. It exists, but it takes effort, and in one case it also means changing your default search engine.

For product managers, this is a good example. If you are building AI into an existing product, the "off" switch is better kept inside the product itself, not in a separate extension. Part of the audience does not want generation, and that is a legitimate request, not a misunderstanding. When the switch is hidden or tied to side conditions, trust suffers more than if it did not exist at all.

Meanwhile: Bing as a Search Engine for Agents

Still, Microsoft is not giving up on AI in search. The focus is simply shifting from people to machines. This summer the company introduced the Web IQ set of APIs built on Bing. It is positioned as a search engine for AI agents, which need structured data rather than a page with ten blue links.

Here is what is known about how it works:

• Web IQ runs on Bing's global index. That is billions of documents in more than 100 languages and markets.
• For semantic search, the system does not compare keywords but projects information into a vector space. It uses an open-source embedding model from Microsoft.
• For speed, Microsoft is extending DiskANN. This technology makes it possible to work with large indexes stored on disk without loading them entirely into memory.

For agent developers, this means Bing now has a competitive offering alongside other search APIs. The value here lies in a ready-made index, vector search and structured output, for which you won't have to build your own crawler. The material includes no public prices, benchmarks or comparisons with competitors, so for now this is a reason to read the documentation, not to migrate.

What to Do About It

If you are simply a Bing user and AI answers get in your way, choose between the extension and "-ai". But remember that the extension will change your default search engine. If you are building a product with AI features, take note: the toggle should be simple, persistent and located where the user actually uses it. If you are building agents, take a look at Web IQ as one source of web data. But first test the quality on your own queries.

The main takeaway: even at a company that has invested heavily in AI search, users demand the right to opt out. It is better to build that demand into a product from the start than to bolt it on later as a workaround.