OpenAI is selling ads inside ChatGPT. Perplexity just killed theirs. Google is hedging. Anthropic ran a Super Bowl ad mocking the whole idea. The way AI platforms make money will determine what they become.

Two weeks ago, OpenAI began showing ads to free-tier ChatGPT users in the United States. This week, we learned who is buying: Albertsons ran Valentine's Day flower promotions inside the chatbot. Williams-Sonoma is putting high-end kitchenware in front of users mid-conversation. Target says traffic from ChatGPT to Target.com has been growing 40 percent month over month.

Three of America's biggest retailers now treat a chatbot as an advertising channel. Meanwhile, the company that pioneered AI search ads just abandoned them entirely.

When we first covered OpenAI's ad launch, Target and its retail media arm Roundel were the sole confirmed partner. The programme has expanded quickly.

Albertsons announced its participation timed to Valentine's Day, running ads that appear when users search for things like "best flowers for Valentine's Day." Williams-Sonoma is testing product placements across its portfolio, including Pottery Barn and West Elm, targeting users as they make purchasing decisions. Adobe is promoting Acrobat Studio and Firefly through its agency partner WPP.

The agency infrastructure is already forming. Omnicom Media has secured placements for more than 30 clients spanning automotive, beauty, retail, and hospitality. WPP and Dentsu are both running campaigns. The minimum buy-in is $200,000, and OpenAI is charging a $60 CPM, three times what Meta charges and on par with Netflix at launch. This is not experimental pricing. It is premium positioning.

The economics behind the decision have not changed since our previous coverage. OpenAI earned $12.7 billion in 2025 but spent roughly $22 billion. With 800 million weekly active users and only five percent converting to paid subscriptions, advertising was not a creative choice. It was arithmetic.

On the same week OpenAI was onboarding retailers, Perplexity was walking in the opposite direction. The contrast could not be sharper.

The AI search company was one of the first to test sponsored answers back in 2024, displaying ads beneath chatbot responses. It has now completely abandoned the strategy. Taz Patel, the company's head of advertising sales, departed in August 2025. Jessica Chan, head of publisher partnerships, confirmed at Advertising Week that Perplexity had stopped accepting new advertisers and had no plans to resume.

The reasoning was blunt. "The challenge with ads is that a user would just start doubting everything," one executive told the Financial Times. "A user needs to believe this is the best possible answer."

Perplexity is now valued at $18 billion and generates $200 million in annualised revenue, the vast majority from subscriptions ranging from $20 to $200 per month. The company is ramping its enterprise sales team, currently just five people, and targeting finance professionals, CEOs, and doctors who will pay a premium for answers they can trust. One executive said the company might "never ever need to do ads."

Then there is Anthropic. The company did not just decline to run ads. It spent millions mocking the entire concept during Super Bowl LX. The campaign, created with agency Mother and directed by Jeff Low, featured the tagline: "Ads are coming to AI. But not to Claude." One spot showed a man speaking to an AI therapist about his relationship with his mother, only to be served a dating app promotion mid-session. The satirical jab at ad-supported AI pushed Claude from No. 41 to No. 7 on the US App Store, with downloads jumping 32 percent in the three days following the broadcast.

Anthropic's business model backs the positioning. More than 80 percent of revenue comes from enterprise contracts, and the company has surpassed $9 billion in annual run-rate revenue in under two years. Though it has left itself the option to revisit the decision, the competitive distance from OpenAI is now deliberate.

Google is playing a more complex game. It is the only major platform running ads in AI-generated search results while keeping its standalone chatbot ad-free.

Ads are already live inside AI Overviews, the summaries that appear above traditional search results. That feature now reaches over two billion users monthly across more than 200 countries. In February, Google introduced a shopping ad format inside AI Mode, its conversational search experience with over 75 million daily active users. Ads appear below AI-generated answers when Google detects commercial intent.

But Gemini, Google's standalone chatbot and direct competitor to ChatGPT, remains ad-free. Dan Taylor, VP of Global Ads, has acknowledged the tension: ads shown too early in conversational AI feel "intrusive" and create "a trust problem." Google has confirmed plans to bring ads to Gemini eventually, but has not committed to a date.

The distinction matters. Google is testing whether you can have ads in the AI layer of search without contaminating the AI product itself. If it works, Google gets both the ad revenue and the trust premium. If users stop distinguishing between AI Overviews and Gemini, the wall collapses. Google also recently introduced its Universal Commerce Protocol, enabling checkout directly within Search and Gemini for participating US retailers including Target, Walmart, and Shopify merchants. The advertising and transaction layers are merging.

What we are watching is not just a business strategy disagreement. It is a structural divergence that will shape what these platforms become.

On one side: OpenAI with 800 million weekly users, a $60 CPM, three of the largest advertising holding companies on board, and analyst projections of $17 billion in ad revenue by year-end. Google with ads already running in AI search to two billion users.

On the other: Perplexity with 100 million users, $200 million in subscription revenue, and a bet that premium users will pay for uncompromised answers. Anthropic with $9 billion in enterprise run-rate revenue, a Super Bowl ad that turned "no ads" into a brand identity, and Claude climbing the App Store charts on the strength of it.

The split mirrors one we have seen before. Cable television fractured into ad-supported free channels and premium subscription services. Streaming repeated the pattern. Now AI is running the same experiment, except the product being shaped by the business model is not entertainment. It is the layer through which hundreds of millions of people are starting to make decisions, get recommendations, and spend money.

The next 12 months will determine whether both models can coexist or whether users and budgets consolidate around one. Several pressure points will force the issue.

For commerce, the immediate question is attribution. Target can measure traffic from ChatGPT, but connecting a conversational recommendation to a completed transaction remains unsolved. Retail media has spent years building closed-loop attribution on platforms like Amazon. Conversational AI has no equivalent infrastructure yet.

For agencies, the signal is clear. Omnicom, WPP, and Dentsu are already allocating client budgets. A self-service ad platform from OpenAI, widely expected by 2027-2028, would bypass them entirely. Asad Awan, who leads monetisation at OpenAI, has already described a future where advertisers prompt ChatGPT directly instead of briefing agencies.

For the ad-free camp, the pressure is on subscription conversion. Anthropic can sustain itself on enterprise contracts. Perplexity, at $200 million in revenue against an $18 billion valuation, needs its premium bet to scale faster.

And for Google, the question is how long the Gemini firewall holds. If AI Mode ads perform well enough, the business case for keeping Gemini clean gets harder to defend every quarter. The company that built the most successful advertising business in history is unlikely to leave a conversational AI product ad-free indefinitely.

If the chatbot recommending your dinner recipe is also selling you the ingredients, does it matter whether it calls that an ad?