HomeResourcesAdvertising and AI
Advertising and AIDo LLMs Read Ads? What the Data Says About Paid vs. Earned Influence on AI Answers
Brands spend billions on advertising, but AI assistants build their answers from somewhere else. Here is what the data shows about where ads fit, and where they do not.
- AI models do not "read" your display, social, or search ads the way they read articles. Ad creative is rarely part of the content they learn from or retrieve.
- In Muck Rack’s May 2026 analysis of more than 25 million links, earned media made up 84% of AI citations. Paid and advertorial content made up just 0.3%.
- Ad spend correlates only weakly with AI visibility (0.215 in Ahrefs’ 75,000-brand study), about a third as strongly as brand web mentions (0.664).
- Ads can still matter, but only indirectly: through the people, coverage, and conversations they create.
The short answer: no, not directly
It is one of the most common questions marketers ask about generative AI: if we are spending heavily on advertising, does ChatGPT, Gemini, or Perplexity notice? The honest answer is that large language models do not read ads in any meaningful sense. To see why, it helps to look at the two ways an AI assistant learns what to say about a brand.
Path 1: training data
Models are trained on huge snapshots of text: articles, reference pages, forums, books, and documentation. Display banners, social ads, and search ads are mostly images, short snippets, or units served by ad platforms rather than pages of text, so they make up very little of what a model learns from.
Path 2: live retrieval
When an assistant searches the web to answer a question, it pulls in pages it can crawl and cite: reviews, news stories, comparison guides, community threads. An ad that appears next to those pages is not the page itself, and assistants cite the content, not the ad slot.
The data on what assistants actually cite backs this up.
What AI actually cites: 84% earned, 0.3% paid
Muck Rack’s What Is AI Reading? report tracks the links that ChatGPT, Claude, and Gemini cite in their answers. Its May 2026 edition analyzed more than 25 million links across 17 industries and found that 84% of citations came from earned media: coverage, reviews, and mentions a brand did not pay for. Paid and advertorial content accounted for just 0.3%.1 Muck Rack notes that the 84% figure has held steady across three reports since July 2025.
View the data
| Citation type | Share |
|---|---|
| Earned media | 84% |
| Paid and advertorial | 0.3% |
| Owned and other (remainder) | 15.7% |
In other words, if you picked 1,000 random citations from AI answers, you would expect to find about three that came from paid placements. The other 997 come from content someone chose to write about a topic, a product, or a brand.
Trace an AI answer back to its sources
To make this concrete, here is a sample AI answer about a fictional product category. Hover over or tap each sentence to see the type of source that typically supports a claim like it. Then look at the sponsored unit beneath the answer.
Notice the pattern. The recommendation itself, which brands get named and why, comes almost entirely from earned sources. The brand’s own site fills in facts like pricing. The ad sits outside the answer entirely.
Ad spend vs. mentions: what correlates with AI visibility
Citations tell you what assistants link to. A second dataset looks at which brand signals line up with being mentioned at all. Ahrefs analyzed 75,000 brands and measured how strongly different factors correlated with how often each brand appeared in Google’s AI Overviews.3
View the data
| Factor | Correlation |
|---|---|
| Branded web mentions | 0.664 |
| Branded search volume | 0.392 |
| Ad traffic | 0.216 |
| Ad cost | 0.215 |
Branded web mentions came out on top at 0.664. Ad traffic (0.216) and ad cost (0.215) sat near the bottom of the list, roughly a third as strong. Spending more on ads, on its own, does not track closely with being named in AI answers.
The middle bar is the interesting one. Branded search volume (0.392) sits between the two, and it is one of the things advertising can genuinely move. That is the first hint of how ads can matter after all.
How ads can still influence AI, indirectly
Ads cannot talk to a model, but they can talk to people, and people create the content models learn from. When a campaign makes someone search for a brand, write a review, start a thread, or pitch a story, it can leave a trail of earned content behind. That trail is what assistants read.
This also explains why two campaigns with identical budgets can have very different effects on AI visibility. A campaign that sparks reviews, coverage, and conversation leaves material for AI to find. A campaign that only drives clicks to a landing page leaves very little.
What about ads inside ChatGPT?
There is a newer wrinkle. In February 2026, OpenAI began testing ads in ChatGPT for logged-in adult users on its Free and Go tiers. OpenAI states plainly that “ads do not influence the answers ChatGPT gives you,” and that ads are labeled as sponsored and visually separated from the answer. Pro, Business, Enterprise, and Education plans do not show ads.2
So even when you can buy space inside an AI assistant, you are buying a placement next to the answer, not a place in it. The organic answer is still built from the same earned sources as before.
- Judge campaigns by the content they create, not just clicks. Track reviews, articles, threads, and branded searches that follow a campaign, not only conversions.
- Brief creative to spark conversation. Claims worth repeating, data worth citing, and offers worth reviewing give people a reason to write about you.
- Treat earned coverage as a media outcome. If 84% of citations are earned, earned media belongs in the same plan as paid media, not a separate silo.
- Keep ad placements and AI answers separate in reporting. Ads inside assistants can drive traffic, but they do not change what the assistant recommends.
Frequently asked questions
Do AI models train on ads?
Very little. Ads are mostly images, short snippets, or units served by ad platforms rather than crawlable pages of text, so they make up a tiny share of training data compared with articles, forums, and reference content.
Do ads in ChatGPT change its answers?
According to OpenAI, no. Its February 2026 announcement says ads do not influence ChatGPT’s answers and are shown labeled and separated from the response.
Can paid content ever be cited by AI?
Occasionally. Muck Rack found paid and advertorial content made up 0.3% of AI citations, so it happens, but rarely compared with earned media at 84%.
If ads do not count, why do some heavy advertisers show up in AI answers?
Usually because large campaigns also generate press coverage, reviews, searches, and discussion. Those earned signals, not the ads themselves, are what AI tools pick up.
Sources
- Muck Rack, “Generative Pulse: Earned Media Consistently Drives AI Citations, Holding at 84%,” What Is AI Reading? report, May 2026. Press release
- OpenAI, “Testing ads in ChatGPT,” February 9, 2026. openai.com
- Louise Linehan and Xibeijia Guan, “An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied),” Ahrefs, May 26, 2025. ahrefs.com
Convert ad spend to mentions in AI answers.
