Will AI companies embrace ad models? Perplexity and Anthropic step away as OpenAI doubles-down
Analysis
Perplexity’s leadership has confirmed there are “no plans to revisit ads” as a revenue source “anytime soon,” following the company’s removal of ads from its interface late last year.
Despite being one of the first to test the waters by adding sponsored content beneath chatbot answers in 2024, the company has since U-turned, arguing future revenue will rely on paid subscriptions to avoid “trust issues with consumers” and the “potential risk of alienating users.”
The move directly contrasts market competitors, which appear to be accelerating in the opposite direction to generate revenue from free users and provide investor confidence.
As reported by Business Insider, a Perplexity executive said: “The challenge with ads is that a user would just start doubting everything…which is why we don’t see it as a fruitful thing to focus on right now.”
This marks a change of heart, as when the San Francisco-based company began testing ads in 2024, an executive highlighted: “Experience has taught us that subscriptions alone do not generate enough revenue to create a sustainable revenue-sharing program.”
Perplexity AI has over 45m monthly active users (MAUs) and has generated an annual recurring revenue (ARR) of $148m.
A look at the other players
Anthropic, which operates Claude, is the only other company to make a public commitment against implementing ads, arguing that, given the sensitive or complex nature of conversations within the chatbot, advertising would “feel incongruous and, in many cases, inappropriate.”
The company has said it plans to support commerce through user-initiated features such as agentic commerce, where Claude acts on a user’s behalf to complete a purchase, but ads would not appear within the chat window, as this would “compromise” what they want it to be — “a clear space to think and work.”
Anthropic has been publicly clear about its stance on the inclusion of ads within chatbots, with its recent rollout of four Super Bowl commercials that mocked an unnamed AI for suddenly pivoting from flattering responses to hard sells.
There has been speculation that the ads refer to OpenAI’s own U-turn with its chatbot, ChatGPT.
The parent company of the popular chatbot recently announced it is testing ads in the US for logged-in adult users on the Free and new subscription, ChatGPT Go, which costs $8 per month.
Despite suggestions from competitor Anthropic, OpenAI has maintained that ads will be labelled, separate from answers and will not influence responses.
Additionally, user data will not be shared with advertising companies.
The move reflects a change in CEO Sam Altman’s stance on ads, where he originally told Harvard Business School in 2024, “ads are a last resort” and “ads-plus-AI is uniquely unsettling,” to his position now.
Arguably, the decision reflects investor pressure for ChatGPT to become profitable.
ChatGPT has 800m weekly active users, but its actual MAU is expected to be under 1bn. OpenAI’s ARR as of June 2025 was $10bn, including revenue from consumer products, ChatGPT business products, and API services. Only 5% of ChatGPT’s WAU are paid subscribers.
Notably, a recent $40bn funding round in March 2025 valued the company at about 30 times the revenue.
Microsoft Copilot has already integrated sponsored content into the experience, and ads appear below responses with explanations of how they relate to the conversation.
The Financial Times reported that Elon Musk told advertisers last summer that the platform plans to introduce ads to Grok.
While Meta AI does not insert ads directly into chatbot responses, Meta has confirmed that interactions with its AI tools are used to personalise content and advertising across its platforms. It has also not publicly ruled out ads within AI conversations.
The trust issue
For users, the inclusion of ads in AI responses may raise questions as to why the ad is appearing and who it benefits.
The main issue is that users should not have to question whether their AI is steering them toward a product rather than offering unbiased advice.
Claire Holubowskyj, senior research analyst, Enders analysis, said: “A key value of chatbots is in how their only incentive is to give the best possible answers to queries — introducing advertising, even if not directly integrated into responses, erodes this by casting doubt on whether serving users or monetisation is motivating recommendations.”
What’s more, the inclusion of ads may direct users towards chatbots without commercial interruptions, potentially leading to a market shape-up.
ChatGPT is the current market leader at 64.5%, but that is a 22 percentage-point decline since the beginning of 2025, according to the Global AI Tracker from SimilarWeb.
The risks of engagement optimisation
Another issue with introducing ads into the revenue model is that it risks incentivising engagement for revenue.
Zoë Hitzig, a former OpenAI researcher, highlighted this issue in a recent opinion piece for The New York Times.
Hitzig outlined that while ads are not inherently problematic, they can “create incentives which reshape systems over time.”
She pointed to the nature of conversations with chatbots, how users typically share personal thoughts and details, and the potential dangers of advertising models built on this.
“Advertising built on that archive creates a potential for manipulating users in ways we don’t have the tools to understand, let alone prevent,” Hitzig wrote.
Hitzig underlined the slippery slope introducing advertising could bring, and despite OpenAI maintaining that ads will not influence answers to queries, Hitzig is not convinced this is a hardline policy — referencing how the company already optimises daily engagement through complementary responses.
Ads allow for access
Despite concerns and clear issues that need to be addressed, advertising revenue enables access for consumers who cannot afford expensive subscriptions.
Holubowskyj said: “How necessary advertising is will depend on a model’s user base: higher-value enterprise subscriptions are the clearest route to profitability, but are only feasible for a handful of tools.
“Most audiences won’t be upsold to higher paid tiers, so for consumer-oriented model providers, advertising will be the main way to monetise reach.”
Top-tier subscriptions for ChatGPT, Gemini and Claude now cost between $200-$250 a month, demonstrating the role advertising may play in keeping access freely available and enabling AI companies to reach profitable business models.
However, a careful balance will need to be struck to ensure certain consumers are not excluded, while also preventing potential manipulation and erosion of user trust.