LLM Ads in 2026: What Brands Need to Know About Advertising in AI Assistants

Last updated 8 October 2026. This market changes week to week, so check the sources linked at the end before you act on a number.

Ads have arrived inside AI assistants. In February 2026 ChatGPT began showing ads to some users in the United States, and by the end of August OpenAI said the business had reached a $1 billion annualised run rate. At the same time, Perplexity gave up on ads and Anthropic promised that Claude would never carry them. For brands, the question is no longer whether LLM ads exist, but whether they are worth buying yet and what to do about the rest of the AI answer.

This guide pulls together what is live, what it costs, what early results show, what consumers think, and what a sensible brand should do next.

Key takeaways

• ChatGPT is the main new ad channel. Ads started in the US in February 2026 for Free and Go users and have since expanded to more than 40 countries, according to OpenAI figures reported by Trending Topics.
• Most AI ad money is not inside chatbots. eMarketer expects over 80% of US AI ad spend in 2026 to sit next to AI-generated content such as Google AI Overviews. Pure chatbot ads are forecast at under $1 billion this year.
• Early performance is modest and hard to measure. A Graphite analysis of Similarweb data put the US click-through rate at 0.73%.
• Trust is the fault line. In an Ipsos survey, 63% of US adults said ads in AI search would reduce their trust, yet most people say they would accept ads in return for free access.
• For brands, being mentioned in the answer matters more than being sponsored next to it. Earn organic visibility first, then test paid placements with a small budget.

What are LLM ads?

LLM ads, sometimes called generative engine advertising, are paid placements shown in or around the answers of large language model products. eMarketer splits them into three groups, which is a useful way to think about the market:

1. AI search-adjacent ads. Ads shown alongside AI-generated answers in a search product, such as Google AI Overviews and AI Mode.
2. AI conversational search ads. Ads in answer engines built around conversation and research.
3. AI chatbot ads. Ads inside a general-purpose assistant such as ChatGPT.

The distinction matters because the money, the buyers and the rules are very different in each group.

Who runs ads in AI assistants today

ChatGPT ads in detail

Timeline and pricing

OpenAI confirmed it would test ads on 16 January 2026 and began showing them in the US the following month. Launch pricing was reported at about $60 per thousand impressions with a $200,000 minimum commitment, roughly three times typical Meta rates. According to The Next Web, that CPM fell to around $25 within ten weeks and OpenAI added cost-per-click bidding at roughly $3 to $5 per click. Reports differ on the exact minimums, but a self-serve Ads Manager now exists with small daily budgets, so the barrier to a test has dropped sharply.

Scale

On 31 August 2026 OpenAI said ChatGPT ads had passed a $1 billion annualised run rate in under 200 days, with tens of thousands of advertisers and availability in more than 40 countries (Trending Topics). Whether that growth continues is an open question, as the forecast section below shows.

Early results

• Click-through rates. Graphite, using Similarweb panel data published on 29 September 2026, found 0.73% in the US, 0.60% to 0.75% in the UK, Canada, Australia and New Zealand, and 1.21% to 1.91% in Brazil, Mexico, Japan and South Korea. The data covers between five and 25 weeks per market (Search Engine Journal).
• Versus search. Adthena, a search intelligence provider, put ChatGPT ad CTR at 0.91% against a 6.4% benchmark for Google search, as reported by Trending Topics.
• Relevance. SE Ranking analysed more than 50,000 prompts and found 14.35% of ads were off-topic for the prompt, and 96.37% of placements came with no brand mention in the answer itself (via eMarketer, as reported in the same article).
• Conversion. Evidence is mixed. Criteo, which analysed 500 US retailers in February 2026, reports that traffic referred from AI platforms converts at about 1.5 times other referral channels (a vendor analysis), while some agencies have reported clicks they could not match in their analytics. OpenAI only launched its conversion pixel and Conversions API in May 2026.

The honest summary: clicks are cheap relative to the launch CPMs, click rates are low compared with search, and measurement is still immature. It is a channel to test, not one to build a plan around.

The money: forecast versus ambition

eMarketer’s June 2026 forecast puts total US AI advertising at $32.03 billion in 2026, rising to $68.25 billion by 2030. More than 80% of 2026 spend sits next to AI-generated content. Standalone chatbot advertising is forecast at under $1 billion this year and just over $5 billion in 2030. OpenAI’s own ambitions are reported to be far higher than eMarketer’s forecast (PPC Land summary of the report).

The takeaway for a marketer is that the largest pool of AI advertising money is still Google. A plan that ignores AI Overviews and AI Mode in favour of chatbot ads would be backwards for most advertisers.

Trust: the real constraint

Putting ads next to answers is risky because people use assistants for advice, not browsing. The data and the industry’s own behaviour both show it:

• In February 2026, Ipsos found that 63% of 1,085 US adults said ads in AI search results would reduce their trust. Only 24% disagreed.
• Forrester research reported by WARC found 83% of respondents would accept ads in exchange for access to free tiers. People tolerate ads. They do not like them.
• Perplexity removed its ads after concluding they made users suspicious of every answer. Anthropic turned its ad-free stance into a marketing message.
• OpenAI’s own principles stress that ads must not influence answers, must be clearly labelled, and that conversations are not sold to advertisers (OpenAI).

Independent verification that answers stay unbiased is not yet available, so the safest assumption is that trust will stay under scrutiny.

Rules and disclosure

Sponsored content must be identifiable, and an advertiser stays responsible for its own claims whatever labelling the platform applies. Disclosure and AI-transparency rules differ by market and are still evolving, so check requirements with a lawyer before running campaigns in regulated categories. This post is not legal advice.

What brands should do now

1. Win the answer before you buy the ad. The SE Ranking finding shows that many paid placements appear with no brand mention in the answer. If assistants do not already know and cite you, sponsorship will not fix that. Publish clear definitions, structured FAQs and facts that others can cite. Our guide to GEO in South East Asia covers the basics.
2. Test with a capped budget. Start with cost-per-click bidding and a daily limit. Judge results on sign-ups or sales, not clicks, and expect gaps between platform reports and your own analytics.
3. Write for the conversation. The best placements match what the person just asked. Be specific, useful and honest. Use exclusions where available, since off-topic placement is common.
4. Keep your brand consistent across every surface. LLM ads, AI Overviews, shopping assistants and your own site all show your brand in different formats and tones. This is where discipline pays: one voice, one visual identity, and a review step before anything goes out. Brand taste is a competitive advantage in a world of generic AI output, as we argued in Why Brand Taste Is the New Marketing Advantage.
5. Respect the red lines. Health, mental health and politics are excluded on ChatGPT, and under-18s do not see ads. Plan your categories and creative accordingly.
6. Do not abandon Google. Most AI ad spend is still near AI search. Make sure your Search and Shopping campaigns are ready for AI Mode.

What to watch next

• Whether CPMs keep falling from the $60 launch price, and whether OpenAI can reach its long-term targets.
• Better measurement and conversion tracking inside ads managers.
• Agentic checkout, where the assistant completes the purchase, and how ads fit into it (Google’s January 2026 announcement).
• Regulation, disclosure rules and any independent audit of whether ads affect answers.

Frequently asked questions

What are LLM ads?

Paid placements in or beside the answers of AI assistants and AI search products, such as ChatGPT, Google AI Mode and Microsoft Copilot.

Does ChatGPT show ads?

Yes. ChatGPT began showing labelled ads in the US in February 2026 to Free and Go users and has expanded to more than 40 countries. OpenAI says ads are not shown to under-18s or near health, mental health or political topics.

Do ads change what an AI assistant says?

OpenAI says its answers are not influenced by advertising, and that ads are separate and labelled. There is no independent audit yet, which is why trust remains the central issue.

How much do ChatGPT ads cost?

Launch pricing was about $60 CPM with a $200,000 minimum. Reports since then describe CPMs near $25, cost-per-click bids of roughly $3 to $5, and a self-serve Ads Manager with small daily budgets. Prices are changing quickly.

Which AI assistants have no ads?

Anthropic has pledged that Claude will stay ad-free, and Perplexity ended its ad program in February 2026.

Should a small brand advertise in ChatGPT now?

Only as a small, measured test. Make sure your brand is already clear and citable in AI answers first, and set a goal you can track.

Sources

• ChatGPT Ads hits $1 billion run rate, but advertisers are not sold. Trending Topics (advertisers, markets, Adthena and SE Ranking figures)
• ChatGPT ad CTR topped 1% in four markets, not in the US. Search Engine Journal, 29 September 2026 (Graphite and Similarweb analysis)
• OpenAI shifts ChatGPT ads to cost-per-click. The Next Web (pricing changes)
• Our approach to advertising and expanding access. OpenAI (principles and restrictions)
• eMarketer says US AI ad spend hits $68bn by 2030. PPC Land, 4 June 2026 (forecast)
• Trust in AI search could drop with ads. Search Engine Journal (Ipsos survey, February 2026)
• Consumers will accept ads in AI chatbots but trust issues remain. WARC (Forrester data)
• Perplexity pulls the plug on ads. Campaign, Anthropic pledges ad-free Claude. Axios, 4 February 2026
• Agentic commerce, AI tools and protocol. Google, Google is bringing ads to AI Mode. TechCrunch, Copilot Showroom ads. Search Engine Land, Meta AI chats and ads. Business Today, Rufus growth. Modern Retail, Ads in Grok. Search Engine Land