--- title: "Are ads coming to ChatGPT and other AI assistants? | Underneath" description: "They are already arriving. Researchers report OpenAI and Google have put ads into their AI products, and ChatGPT ads reached 31 European markets in August." canonical: "https://underneath.agency/resources/are-ads-coming-to-ai-assistants" published: 2026-10-07 updated: 2026-10-08 publisher: "Underneath (https://underneath.agency/agent)" entity: "https://underneath.agency/.well-known/entity.json" --- Guide · AI search # Are ads coming to ChatGPT and other AI assistants? They are already arriving. Researchers report that OpenAI and Google have integrated advertising into their AI products, while Anthropic has stayed ad-free. How far ads will spread inside the answers themselves is still open, and the research suggests competition may hold them back. ## The short version 1. A 2026 Carnegie Mellon paper reports that OpenAI and Google have integrated advertising into their AI products, while Anthropic remains ad-free ([Zhang and colleagues, 2026](https://arxiv.org/abs/2603.29071)). 2. In August, OpenAI expanded ChatGPT advertising to 31 European markets, according to a September 2026 audit that cites other work ([Uberti-Bona Marin and colleagues, 2026](https://arxiv.org/abs/2609.18729)). 3. In a 40-day audit of Google’s AI Overviews, 2.16% of results pages with an Overview also carried a Google sponsored ad, and no ads appeared inside the Overview itself ([Xu, Iqbal and Montgomery, 2026](https://arxiv.org/abs/2605.14021)). 4. US search advertising was a $102.9 billion business in 2024, nearly 40% of all US internet ad revenue, which is the revenue AI answers put at stake ([Khosravi and Yoganarasimhan, 2026](https://arxiv.org/abs/2602.18455)). ## Have AI assistants already started showing ads? Yes. By 2026, researchers describe ads in OpenAI’s and Google’s AI products as a fact, not a forecast. In mid-2025, [Erickson](https://arxiv.org/abs/2506.06447) wrote that it “may only be a matter of time” before conversational search carries ads, noting OpenAI was considering it. That was a prediction. By March 2026, [Zhang and colleagues](https://arxiv.org/abs/2603.29071) at Carnegie Mellon described OpenAI and Google as having integrated advertising into their AI products, with Anthropic strictly ad-free. The spread continued. [Uberti-Bona Marin and colleagues](https://arxiv.org/abs/2609.18729) report that in August, OpenAI expanded ChatGPT advertising to 31 European markets. They cite other work suggesting these ads are most likely to show on product recommendation questions, in 10 to 14% of sessions. That estimate is secondhand, and the paper does not describe the ad format. On Google’s side, [Khosravi and Yoganarasimhan](https://arxiv.org/abs/2602.18455) note that Google had expanded ad placements in AI Overviews (the AI summary at the top of Google’s results) across devices and countries by late 2025. Google’s AI Mode did not yet show ads when [Wang and colleagues](https://arxiv.org/abs/2608.18352) ran their 1,100-person experiment, and the authors expect that to change. ## Why do AI companies want ads at all? Because search advertising is enormous, AI answers are expensive to produce, and buying questions are common. US search advertising generated $102.9 billion in 2024, nearly 40% of all US internet ad revenue. AI answers threaten that model: when the answer appears on the page, fewer people click through to sites that carry ads. Zhang and colleagues add that AI engines face large running costs for every answer, which pushes them toward either ads or paid subscriptions. Buying questions are already a real share of AI use. The largest published measurement of ChatGPT use from 2025 found product and service recommendations made up roughly 2% of conversations, as reported by Uberti-Bona Marin and colleagues. ## Will ads take over AI answers? Not necessarily: an economic model suggests competition between AI engines limits how many ads each can show. [Zhang and colleagues](https://arxiv.org/abs/2603.29071) built a model of an AI engine choosing, question by question, between an answer with ads and one without. They then simulated markets of 500 users over 20 periods, in 160 runs under different conditions. Their conclusions: - Ads pay only when the immediate ad revenue beats the long-term value of keeping a user happy and moving them toward a paid plan. - Ad-heavy policies earn more early but shrink the user base and slow subscriptions. - When rival AI engines are strong, the best policy shifts further toward ad-free answers, because users can switch. Real users do switch. In a field experiment with 1,100 people, forcing searches into Google’s AI Mode raised use of competing search engines by 11.2% ([Wang and colleagues, 2026](https://arxiv.org/abs/2608.18352)). This is a theoretical model with simulated users, not data from any real AI engine. It shows the pressure, not the outcome. ## How do ads sit next to Google’s AI answers today? Mostly beside them, not inside them, in the one large audit we found. [Xu, Iqbal and Montgomery](https://arxiv.org/abs/2605.14021) captured 7,583 results pages with an AI Overview over 40 days in spring 2026: - 2.16% also showed at least one Google sponsored ad on the same page. - 0.51% placed a sponsored ad above the Overview. - No sponsored ads appeared inside the Overview itself. - 50.63% of the pages the Overviews cited ran visible display ads of their own. The authors note their searches were mostly trending, informational ones, where ads are rarer. Commercial searches are likely to carry more ads. How often Overviews appear on commercial searches is measured in [our study of 1,248 US searches](https://underneath.agency/research/ai-overviews-frequency-study). The last point matters for publishers: more than half the cited pages depend on ad-supported visits that an AI answer may replace. ## Would buyers be able to tell an ad from an answer? Not reliably, if ads are blended into the answer, and the research warns that is the risky case. Erickson distinguishes clearly labeled banner ads from [“native” ads written into the answer](https://underneath.agency/resources/risks-of-native-ads-in-ai-answers) itself. Erickson cites earlier research finding users may not notice ads woven into conversational answers unless they are clearly disclosed. The paper calls the danger the “fake friend dilemma”: a user trusts the assistant as a helper while it serves other interests. Presentation already shapes trust. In a September 2026 audit of 1,536 responses, ChatGPT expressed a first-person preference (“my pick would be…”) in 79% of product recommendations, against 7% for Gemini and 2% for AI Overviews. In a separate experiment with 4,927 US adults, adding reference links raised trust in AI answers even when the links were invalid ([Li and Aral, 2025](https://arxiv.org/abs/2504.06435)). Regulators are moving. Uberti-Bona Marin and colleagues report that on 31 August 2026, the European Commission designated ChatGPT a Very Large Online Search Engine, bringing audit and consumer protection duties. [Wen and colleagues](https://arxiv.org/abs/2606.12439) argue US-style rules requiring “Ad” or “Sponsored” [labels should extend to AI answers](https://underneath.agency/resources/should-ai-search-ads-be-labeled). ## What should you do about it? Plan for a paid channel inside AI answers, but keep investing in the earned visibility ads cannot replace. 1. **Assign an owner to track AI ad products.** Watch for self-serve ad programs from OpenAI, Google and others, including where they run and what formats they allow. 2. **Keep earned and paid AI visibility separate in reporting.** Once ads appear beside answers, a mention you earned and a placement you bought are different results. 3. **Protect your organic presence in answers.** Modeling suggests ads will be limited where rival engines compete, and Anthropic remains ad-free, so earned visibility will still carry much of the load. We weigh that bet in [planning AI search for five years](https://underneath.agency/resources/ai-search-strategy-next-five-years). 4. **Insist on clear labels if you buy.** Disguised placements risk buyer trust and regulatory attention. 5. **If you publish content, watch traffic.** More than half of pages cited in AI Overviews run display ads, so lost clicks hit ad revenue. If you want help building earned visibility in AI answers, see our [generative engine optimization service](https://underneath.agency/services/generative-engine-optimization). ## What does the research not tell us yet? The research confirms ads have arrived in some AI products but says little about how well they work. - No paper we reviewed measures click or conversion rates for ads inside AI assistants. - The European rollout figures and session estimate are reported secondhand, not measured by the researchers. - The economic model that predicts limits on ads uses simulated users and assumptions, not platform data. - The AI Overview ad audit covered mostly trending searches, so ad frequency on commercial searches is unmeasured. - Whether ad-supported answers become less accurate or less balanced has not been tested. ## Frequently asked questions ### Does ChatGPT show ads? According to researchers, yes in some markets. A September 2026 audit reports OpenAI expanded ChatGPT advertising to 31 European markets in August. ### Are there ads in Google’s AI Overviews? Next to them, at times. A 40-day audit found 2.16% of pages with an Overview also showed a sponsored ad, and none inside the Overview itself. ### Will AI assistants become full of ads like search results? Not necessarily. A Carnegie Mellon model, run over 160 simulations, found that strong rival AI engines push the best policy toward ad-free answers. ### Is Claude ad-free? As of March 2026, researchers described Anthropic as remaining strictly ad-free while OpenAI and Google had integrated advertising. ## Sources - Erickson (2025), [Fake Friends and Sponsored Ads: The Risks of Advertising in Conversational Search](https://arxiv.org/abs/2506.06447), arXiv:2506.06447. - Zhang, Jiao, Li and Xiong (2026), [An Economic Framework for Generative Engines: Advertising or Subscription?](https://arxiv.org/abs/2603.29071), arXiv:2603.29071. - Uberti-Bona Marin and colleagues (2026), ["If I Had to Buy Just ONE: Galaxy S26 Ultra": Auditing AI-Generated Product Recommendations](https://arxiv.org/abs/2609.18729), arXiv:2609.18729. - Xu, Iqbal and Montgomery (2026), [Measuring Google AI Overviews: Activation, Source Quality, Claim Fidelity, and Publisher Impact](https://arxiv.org/abs/2605.14021), arXiv:2605.14021. - Khosravi and Yoganarasimhan (2026), [Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia](https://arxiv.org/abs/2602.18455), arXiv:2602.18455. - Wang, Gleason, Bart, Wilson and Metaxa (2026), [AI in Search Reduces Publisher Referrals Without Improving User Experience: Experimental Evidence](https://arxiv.org/abs/2608.18352), arXiv:2608.18352. - Li and Aral (2025), [Human Trust in AI Search: A Large-Scale Experiment](https://arxiv.org/abs/2504.06435), arXiv:2504.06435. - Wen and colleagues (2026), [Position: Generative Engine Optimization Creates Underexamined Risks, Governance Must Target Concentration, Disclosure, and Academic Blind Spots](https://arxiv.org/abs/2606.12439), arXiv:2606.12439. --- This is the Markdown twin of https://underneath.agency/resources/are-ads-coming-to-ai-assistants. The HTML page is canonical. Publisher: Underneath, https://underneath.agency/agent. Site index: https://underneath.agency/llms.txt.