Why Does an AI Assistant Get Your Pricing Wrong?

Why does ChatGPT quote a price you stopped charging a year ago?

Because AI assistants do not read only your pricing page. They piece together an answer from old copies, review sites, comparison posts, and your own outdated blog articles. If any of those sources carries an old number, or your page states the price in a confusing way, the assistant can repeat the wrong figure with full confidence.

This is a common AI visibility problem on B2B SaaS sites. A founder asks ChatGPT, Perplexity, or Gemini what their product costs. The answer is wrong. Sometimes it is last year's price. Sometimes it is a plan that no longer exists. Sometimes it is a competitor's price attached to their name.

A wrong price in an AI answer is not a small issue. It can scare off a buyer who would have fit, or set an expectation your sales team has to walk back. The good news is that most causes are fixable on your side. Here is how I diagnose them.

How do AI assistants decide what your product costs?

They combine whatever sources they can retrieve or remember. Some answers come from a live web search at the moment of the question. Others lean on what the model learned during training. Most blend both. Each source can carry a different price, and the assistant has to pick one or reconcile them.

When an assistant searches live, it pulls a handful of pages and quotes passages from them. Your pricing page might be one. A G2 listing, a Reddit thread, or a "best tools" roundup might be others. The model then writes one answer from all of them, and it does not always know which source is newest.

When an assistant answers from training data, the problem is different. The model may have learned your pricing months ago. If you changed prices since then, it may repeat the old figure until it searches again or gets retrained. You cannot edit that memory directly. You can only make the current price easy to find when it does search.

Is your own site sending mixed signals?

Very often, yes. A frequent cause is old prices on the company's own website. A launch post from two years ago, an FAQ that was never updated, a help article with a screenshot of the old plan grid. To an AI assistant, those are all official pages from you, and they disagree.

Run a quick check. Search your own site for your old plan names and old price points. Include blog posts, help docs, case studies, and landing pages built for past campaigns. Every hit is a page that can feed an assistant the wrong answer.

The fix is boring but effective. Update the page, add a clear note that pricing changed with a link to the current page, or retire the page with a redirect if it no longer serves anyone. I covered why assistants often cite the stale version in why AI answers cite an old version of your page.

Can AI assistants actually read your pricing page?

Not always. If prices load through JavaScript after the page renders, sit behind a toggle, or appear only inside an image, some crawlers and retrieval systems may miss them. The safest approach is to put every price, plan name, and billing unit in plain HTML text that is present when the page first loads.

Monthly and annual toggles are a classic trap. The default view might show annual prices, while the monthly prices only appear after a click. An assistant that reads the default view may report the annual price as the monthly one, or mix them. I wrote about this in detail in how AI assistants read a pricing page with a toggle.

A simple test helps. Load your pricing page with JavaScript turned off, or view the raw page source. If you cannot see the prices there, assume some AI systems cannot either. Then add a plain text summary of each plan to the page.

Why do review sites and comparison posts matter so much?

Because assistants often trust third-party pages as much as yours, and those pages update on their own schedule. A comparison article written last spring may still list your old price. A review platform may show a plan you retired. When the assistant sees three outside sources agree on an old number, it may side with them.

Start by listing the third-party pages that rank or get cited for your brand plus "pricing." Check each one. For review platforms where you have a vendor profile, update the pricing fields yourself. For articles, email the author with the new figures and a link. Many will update if you make it easy.

This is slow, unglamorous work. It is also some of the highest leverage AI visibility work you can do for pricing, because it removes the conflicting sources instead of trying to outshout them.

Does how you phrase prices change what AI repeats?

Yes. Ambiguous phrasing invites wrong summaries. "From $49" with no unit could mean per user, per month, or per year. Write prices as complete statements: "$49 per user per month, billed annually." Spell out what is included and what costs extra. An assistant can only repeat clearly what you state clearly.

The same goes for plan names. If your plans are called Starter, Growth, and Scale, use those exact names everywhere: the pricing page, the docs, the sales deck, the help center. If the help center still says "Pro," an assistant may invent a plan that mixes the two.

Currency matters too, if you sell globally. State the currency next to every price. A bare "$" can be read as US, Canadian, or Australian dollars depending on context, and the assistant will guess.

What if your pricing says "contact sales"?

Then the assistant has nothing official to quote and will fill the gap with whatever it finds. That is often a guess from a review site or a forum thread. If you cannot publish exact prices, publish a starting point, a typical range, or the factors that drive price, in plain words.

Even a sentence like "Most teams of 20 to 50 users pay between X and Y per year" gives an assistant something better than a rumor. You stay in control of the range, and buyers get a sense of fit before they book a call.

If you truly cannot share any figure, at least explain how pricing works. Per seat or per usage. Annual or monthly. Minimum contract length. That context reduces the chance of a wildly wrong answer. I went deeper on this trade-off in how AI assistants handle contact sales pricing.

Does Google say anything about optimizing for AI features?

Google published a guide in May 2026 on optimizing websites for generative AI features in Search. According to the Search Central blog post announcing it, the guide covers "valuable, unique, non-commodity content" and explains why "SEO best practices remain relevant and foundational" to success with those AI features.

The same announcement says the guide includes a section mythbusting common "AEO/GEO" misconceptions. The practical reading for pricing is straightforward. There is no secret tag that forces an AI answer to use your price. Clear, crawlable, consistent pages are still the foundation.

That matches how I approach pricing audits. The fixes worth making first are the ones that would also help a human buyer: one clear pricing page, consistent plan names, no stale pages, and plain statements of what things cost.

How do you monitor pricing accuracy in AI answers?

Ask the main assistants the same pricing questions every month and log the answers. Use the questions buyers actually ask: "How much does [product] cost?", "What is the cheapest [product] plan?", and "[product] vs [competitor] pricing." Note which answers are wrong and which sources each assistant cites.

The cited sources are the valuable part. When an answer is wrong, the citation usually points straight at the page causing it. That turns a vague problem into a specific task: update this help article, email this author, fix this toggle.

Repeat the check after every price change. Pricing changes are when AI answers drift the most, because old sources outnumber the new one for a while. Expect a lag, and keep correcting sources until the answers catch up.

What should you do next?

Today, ask ChatGPT, Perplexity, and Gemini what your product costs. Write down each answer and its sources. Then search your own site for old plan names and prices. Those two checks will show you, within an hour, whether the problem starts on your site or somewhere else.

Fix your own pages first, because you control them. Then put prices in plain text, state units and currency, and work through the third-party pages one by one.

If you want help auditing how AI assistants describe your pricing, or rebuilding a pricing page that both buyers and AI systems can read, reach out. I build websites and the automations behind them for lead generation, and getting the basics right here protects every lead that starts with an AI question. Let's chat.