AI tool businesses are everywhere in 2026, and many of them are for sale. A simple tool wrapped around a large language model can grow fast, earn subscription revenue, and look like an easy acquisition. But the same business can also lose its edge overnight when a model provider changes prices, releases a competing feature, or cuts off access.

The first number to examine is the gross margin after model costs. Many AI tools report revenue proudly and hide the API bill in a footnote. Ask for monthly model and infrastructure costs for at least twelve months, then calculate what is left after those costs. A tool with 40 percent margins after model costs is a very different business from one with 90 percent.

Watch how model costs move as usage grows. Some tools become cheaper per user as they optimise prompts and cache results, while others become more expensive because heavy users drive the bill. Ask for cost per active user over time, and be suspicious of any business where that cost is rising faster than revenue per user.

The moat question matters more for AI tools than for most software. If the product is a thin wrapper around a public model with no unique data, workflow, or distribution, a competitor can copy it in a weekend. Look for something defensible: proprietary data, a loyal community, a distribution channel, or a workflow that users have built their work around.

Customer retention tells you whether the moat is real. Check monthly churn and, more importantly, why customers leave. If users churn after one month because they got what they needed, the tool may be a one-time utility, not a subscription business. If they stay for a year and expand usage, the product has solved a recurring problem.

Ask what happens if the main model provider changes its terms. Does the tool depend on a single provider, or can it switch models without rebuilding? Can it run on a cheaper model for most tasks and a stronger one only when needed? Businesses with model flexibility survive provider changes; single-model businesses live at someone else's mercy.

Intellectual property is often thinner than sellers claim. The prompts, the interface, and the brand may be original, but the underlying model is not owned by the seller. Make sure the purchase agreement clearly transfers the prompts, the code, the user data rights, and any custom fine-tuning, and check whether any of it depends on licences that do not transfer.

Pricing power is another test. Has the business ever raised prices, and what happened? AI tools that cannot raise prices without losing most customers are often competing on price alone, which is a race to the bottom as model costs fall and competitors multiply. Tools with pricing power usually have a workflow lock-in or a professional audience.

Look at how the business gets customers, because AI tools often depend on one fragile channel. A tool that grows through search, a marketplace, or a community is more durable than one that depends on a single viral post or a paid campaign that stops working the moment spending stops. Ask for traffic and signup sources by month.

The team question is simple but decisive: who keeps the product working if the model provider ships a breaking change tomorrow? If the answer is only the founder, and the founder is leaving, you need a transition plan and a developer you trust before closing. AI products break in new ways, and someone has to be there to fix them.

Value an AI tool on its current, model-cost-adjusted profit, not on its growth story. Use a lower multiple than you would for a traditional SaaS business unless the moat is genuinely strong, and consider an earnout tied to margin retention so the seller shares the risk of rising model costs.

AI tools can be wonderful acquisitions for buyers who understand the technology and the costs. They are dangerous for buyers who see only the revenue chart and assume the margins will take care of themselves. The difference between those two buyers is a few hours of honest diligence.

Buy the margin, the moat, and the flexibility, not the hype. If all three check out, an AI tool business can be a strong addition to your portfolio. If any one of them fails, walk away and wait for a business whose value you can actually defend.