Ads and sponsored answers in AI chats: how to spot them and get a straight recommendation

Short answer

Yes. ChatGPT, Google's AI Mode and AI Overviews, and Microsoft Copilot can all place ads in or next to an answer (October 2026). The vendors say ads are labelled and kept apart from the answer, and OpenAI says they do not influence ChatGPT's answers. The harder case is an unlabelled answer steered by how pages are written, which research shows is possible. So ask for criteria and sources first, then check the pick outside the chat. The steps below are our suggestion.

For whether to use an AI answer or a search at all, see AI answers or search results: when to use which.

Where the ads appear (October 2026)

• ChatGPT: Free and Go plans only. "Ads are clearly labeled as sponsored and visually separated from ChatGPT's response." They "can appear below the end of a response". None for accounts identified as under 18, near "sensitive or regulated topics, including personal health, mental health, or politics", or in Temporary Chats.
• Google AI Overviews and AI Mode: ads can show "above, below or within AI Overviews", and in AI Mode "may appear below and integrated into AI Mode responses", labelled "Sponsored". One format places an ad in an AI Mode list of recommendations as a "Highlighted Answer", so read every item in a list.
• Microsoft Copilot: ads "are displayed in Copilot's responses", shown "below Copilot's organic response" (2024). The label is not named.
• Shopping results in ChatGPT: OpenAI says "Product results are selected independently by ChatGPT and are not ads, nor influenced by any OpenAI partnerships." Sellers of the same product "are ranked based on factors like availability, price, quality, and whether they are the maker", and some products can be bought inside ChatGPT. Treat the seller order as unverified.

OpenAI says "Ads do not influence ChatGPT's answers." Google and Microsoft do not say either way.

Steered without an ad

An answer can lean towards a product with no ad in it. Adding "a carefully crafted message" to a product page could "significantly increase its likelihood of being listed as the LLM's top recommendation" (LLM means large language model, the kind of AI behind these chats) in a test with "a catalog of fictitious coffee machines" (Kumar and Lakkaraju, 2024, preprint). The same kind of attack worked on "production LLM search engines (Bing and Perplexity)" (Nestaas, Debenedetti and Tramèr, ICLR 2025). Neither measures how often this happens, only that it can. That is why the steps below check every pick outside the chat.

Spot a sponsored unit (1 minute)

Ask for a straight recommendation

I want a recommendation for <ITEM>. My criteria, most important first: <CRITERIA>.
1. Explain how you will weigh these criteria before naming any product.
2. Give up to 3 options. For each, link the source for every claim, and say whether
   that source is the maker, a retailer, an independent test or review, or a page
   that earns money from links.
3. Say if any option or link is sponsored, an ad, or an affiliate link, or say
   that you cannot tell.
4. Tell me which claims you could not check.

Replace the following:

• <ITEM>: such as a kettle for a small kitchen.
• <CRITERIA>: such as quiet, under 1.5 litres, easy to clean, under 40 (illustrative).

Then check outside the chat (about 5 minutes, our estimate):

1. Open the seller's or maker's own page for price, specs and warranty. If it differs from the answer, trust the page.
2. Read one independent review that does not earn from links.
3. Ask again in a fresh chat or another assistant with the same criteria. A pick that appears only once deserves a harder look.

Worked example (illustrative)

A reader on a free ChatGPT plan asks for a kettle. Under the answer sits a box marked "Sponsored" for a brand the answer did not mention. They ignore the box and run the prompt above (the text to paste into the chat) with quiet, under 1.5 litres, easy to clean, under 40.

• The answer weighs quietness first, then names three kettles. Two cite an independent test; the third cites only a retailer page, and the assistant says it cannot tell whether that link earns money.
• On the maker's page, the third kettle is 1.7 litres, so it fails a criterion. They drop it.
• A fresh chat with the same criteria names one of the first two again. They read one independent review of it, check the price on the maker's page, and buy it.

Total time: about 10 minutes. The sponsored box never entered the decision.

Limit ad personalization (5 minutes)

These settings change which ads you see, not whether you see ads (October 2026).

• ChatGPT: Settings > Ad Controls. Turn off Personalize ads, or see why an ad was shown. With it off, "you will still see ads based on the context of your current chat thread". Plus or Pro has no ads, and Free users can choose "a free, no-ads option with additional limits (fewer messages per day and no access to tools like images)".
• Google: in My Ad Center, "Turn on or off personalized ads on Google". With it off, ads still follow "contextual factors like the time of day" and your active searches.
• Microsoft Copilot: in your Microsoft account's personalized ad settings, switch off "See ads that interest you". The personalization setting in Copilot "does not control whether you receive personalized ads". Microsoft's help page says it describes the older Copilot app, so check the setting in the app you use.

For the wider quarterly settings check, see Privacy settings pass: 15 minutes, once a quarter.

How to tell it worked

• You can point to the label on any ad you saw.
• Every option has a source you opened, and you know which sources are makers, retailers, independent reviews or link earners.
• The answer said whether anything was sponsored or affiliate, or said it could not tell.

Sources

1. OpenAI. Ads in ChatGPT (help), October 2026; Our approach to advertising, January 2026. Vendor pages. Link, Link
2. OpenAI. Shopping with ChatGPT Search (help), October 2026, read 2026-10-18. Vendor page. Link
3. Google. About ads and AI Overviews (Ads Help); AI in Search posts, May 2025 and May 2026. Vendor pages. Link, Link, Link
4. Google. My Ad Center Help, October 2026. Vendor pages. Link, Link
5. Microsoft. About ads in Copilot, June 2026; ads announcement, October 2024; Copilot privacy controls, October 2026. Vendor pages. Link, Link, Link
6. Kumar A, Lakkaraju H. Manipulating Large Language Models to Increase Product Visibility, 2024. Preprint. Link
7. Nestaas F, Debenedetti E, Tramèr F. Adversarial Search Engine Optimization for Large Language Models. ICLR 2025. Peer-reviewed conference paper. Link

Changes

1. v1.010 Oct 2026first version