Perplexity Ditched Ads to Keep Your Trust

Perplexity is dropping ads while ChatGPT adds them. Anthropic mocked AI ads at the Super Bowl. The AI business model divide is the biggest strategic split in the industry.

Written by
Richard Batt
Published
24 February 2026
Reading time
6 minutes
Share on LinkedIn

Key takeaways

  • The Business Model Divide Is Now Impossible to Ignore, apply this before building anything.
  • Follow the Money: How Ad-Supported AI Tools Change Their Behaviour.
  • Why Enterprise Buyers Should Care More Than Anyone and what to do about it.
  • The Framework for Evaluating AI Vendors: A Practical Checklist, apply this before building anything.

Last month I watched Anthropic’s Super Bowl advertisement with genuine fascination. It was not a typical tech company ad. Instead of celebrating their Claude AI model, they showed a series of satirical scenes: a family being interrupted by AI notifications, an office where every tool now displays ads, a world where artificial intelligence has become the primary way companies manipulate user attention.

The message was unmistakable. And then Sam Altman, CEO of OpenAI, responded on social media with what amounted to a public rebuke. He defended ChatGPT’s recent decision to begin showing advertisements to users. His argument: ads are how you keep products free or cheap.

What I am witnessing is not a marketing spat between executives. It is a fundamental business model war that will reshape how we interact with AI for the next decade. And I can tell you, based on watching this play out across 120+ consulting projects, which side your business should pay very close attention to.

The Business Model Divide Is Now Impossible to Ignore

Perplexity, the search-focused AI startup, has made a quiet but significant choice. They are phasing out their free advertising-supported tier and moving toward subscriptions and enterprise deals. Their Series B fundraising deck, the details of which circulated through investor networks, makes clear that advertising is not their long-term strategy.

OpenAI has done the opposite. ChatGPT’s free tier now displays advertisements. The ChatGPT Plus subscription (US$20 per month) remains ad-free, which is telling on its own. The incentive structure is inverted: OpenAI is now motivated to make the free experience slightly worse (ad-laden) so that users upgrade to paid tiers.

Anthropic is not just talking about avoiding ads. They are positioning their entire company around the philosophy that when your AI tool contains advertising, its incentives become misaligned with your interests. The Super Bowl ad was expensive. But it was a bet that users would pay more for an ad-free experience if they truly believed the alternative we compromised.

Practical tip: When evaluating any AI tool for your business, whether for customer service, content generation, research, or analysis, ask yourself one question: who does the AI vendor make money from? If the answer is advertisers rather than you, that company has an incentive to optimize for advertiser goals, not your goals.

Follow the Money: How Ad-Supported AI Tools Change Their Behaviour

This is not theoretical. I have observed this dynamic across consulting engagements where organisations deployed AI tools with different business models, and the effects were measurable.

One manufacturing client I advised in 2024 was using an AI-powered supplier recommendation tool that relied on advertising revenue. Over the course of six months, I noticed something subtle: the tool began recommending suppliers who paid for sponsored placement more often than its original training data suggested was optimal. The change was not overt. It was gradual. But when we tracked which suppliers we recommended, and then cross-referenced against that vendor’s sponsorship status with the platform, the pattern emerged. The tool was not broken. Its incentives had just shifted slightly.

A professional services firm I worked with in 2025 adopted an ad-supported AI assistant for their client communication workflows. The assistant would occasionally recommend external tools, services, or resources within client conversations. After three months, the firm’s partners noticed that the recommendations seemed to favour services that advertised on the platform. Again, this was not obvious manipulation. It was subtle. But when your business model depends on advertising revenue, that bias enters the system.

This is what Anthropic is betting against. They are saying: we do not want that incentive structure in our product. If we take advertising, we are compromised.

Sam Altman’s counterargument, that ads allow cheaper access, is not wrong. Free or low-cost products have historically sustained through advertising. But there is something more interesting happening. Users are increasingly willing to pay for ad-free experiences. Look at the success of subscription-based products: Spotify Premium, Netflix ad-free tiers, YouTube Premium. Users will pay to avoid ads and the incentive misalignment they represent.

Why Enterprise Buyers Should Care More Than Anyone

If you are evaluating AI tools for your organisation, this business model divide should be one of your top three evaluation criteria. Here is why.

When you deploy an ad-supported AI tool into your business, you are not just accepting ads. You are accepting subtle shifts in the tool’s recommendations, priorities, and outputs. In some domains, creative writing, brainstorming, this not matter. In others, supplier selection, hiring recommendations, financial analysis, customer segmentation, even a 2% bias introduced by advertising incentives can compound into significant business decisions.

A financial services client I advised last year was considering an ad-supported platform for regulatory research and compliance analysis. I advised against it, not because the tool was technically inferior, but because financial regulators are increasingly scrutinising AI decision-making. If your AI tool’s recommendations have even a whisper of bias introduced by its business model, that becomes a regulatory liability.

Enterprise contracts also reveal something important. When Perplexity and Anthropic sell to enterprise customers, they sell subscriptions and white-label arrangements. OpenAI offers the same. But the economics are different. The enterprise customer becomes the revenue source, not a venue for advertising. That realigns incentives entirely.

The Framework for Evaluating AI Vendors: A Practical Checklist

Here is what I now recommend to every client evaluating AI tools for business-critical use:

Question 1: Where does the vendor make money? Is it from users (subscription), from enterprises (licensing), from advertisers, or from data? If advertisers are in that revenue mix, you now know there is a potential incentive conflict.

Question 2: Has the vendor’s output changed since they introduced advertising? If they recently added ads, compare historical outputs to current outputs on a sample of test prompts. This is easy to do and remarkably revealing.

Question 3: Can the vendor transparently explain their incentive structure? If they hesitate or cannot articulate how they make money, that is a red flag.

Question 4: Are there contractual guarantees about the AI tool’s behaviour? Subscription-based vendors can offer SLA guarantees about output quality because they are not serving competing interests. Ad-supported vendors cannot make those same guarantees.

Question 5: What does the regulatory environment require? If your industry is regulated (finance, healthcare, legal), check whether regulators have issued guidance on acceptable AI vendor business models. Some are beginning to.

I have been using Claude (made by Anthropic) extensively across my consulting practice. I have also used ChatGPT. I notice that Anthropic has been transparent about their position: they are funded by venture capital and customers, not advertisers. That transparency, combined with their Super Bowl gambit, tells me they are betting that their business model is defensible in the long run.

OpenAI is betting differently. They are betting that ubiquity and network effects will matter more than incentive alignment. Both could be right. But the bet you make about which side is correct will shape which tools you deploy in your business for the next five years.

This business model war is not about which AI model is technically superior. It is about which company can maintain user trust while sustaining revenue growth. And I can tell you that in every consulting engagement where I am helping a client choose an AI tool, the business model question has become the deciding factor, more important than model performance, more important than price, more important than features.

Tagged: AI Strategy, Business

Richard Batt
Richard Batt

I build automations, AI steps and custom software for businesses with 5 to 200 staff. 120+ projects across 15+ industries, from Middlesbrough.

About meSee the work

More on ai strategy

All writing

Need a workflow built? Tell me what is slowing your team down.

One paragraph is enough. I reply personally.

Bring me your worst workflow