Ask ChatGPT or Gemini "how much does [product] cost?" about almost any SaaS tool and watch where the answer comes from. Nine times out of ten, the model is paraphrasing the pricing page. Not the homepage, not the blog, not the launch announcement. The pricing page.
That page has quietly become one of the highest-leverage pieces of text on your site, because it is one of the few pages that reads as fact rather than persuasion. Models treat it as reference material. Buyers treat it as the moment of truth. And most companies still write it as an afterthought: a grid of tiers, a few checkmarks, and a "contact us" button.
Here is why that is now a real problem, and what to do about it.
## Why assistants lean on pricing pages
When someone asks an assistant "is X worth it?" or "what does X cost for a team of ten?", the model needs concrete, structured facts: plan names, prices, limits, who each plan is for. Your homepage offers positioning language. Your blog offers narratives. Your pricing page is the one place where the facts live in a compact, unambiguous form.
Assistants strongly prefer sources like that. Clear tier names, explicit numbers, short definitions of each plan's audience: this is exactly the shape of text a language model can lift into an answer with low risk of misquoting you.
The reverse is also true. If your pricing page is vague, the model goes looking elsewhere, and "elsewhere" means review sites, old Reddit threads, a competitor's comparison page, or a directory listing you forgot about. Each of those is someone else's version of your pricing.
## The mistakes that produce wrong answers
A few patterns reliably produce bad AI answers:
**"Contact us for pricing" with nothing else.** The model has nothing to quote, so it either says "pricing is not public" (which reads as "expensive") or it finds a third-party estimate, often wrong, often years old. Even a starting price or a range gives the model something accurate to say.
**Tier names without tier definitions.** "Starter, Growth, Enterprise" tells a model nothing about who belongs where. One plain sentence per tier, "for solo founders", "for teams up to 25", is what turns into the assistant's recommendation logic.
**Prices buried in images or interactive widgets.** If the number only exists inside a graphic or a toggleable monthly/annual switcher rendered by JavaScript, many crawlers never see it. The text version has to exist in the page itself.
**Outdated pages left live.** Old pricing URLs, archived launch posts, and cached directory listings all remain quotable. When they conflict with your current page, assistants average the mess, and the answer usually mentions the old, lower price. That is a painful conversation for your sales team to inherit.
**No answer to "is there a free plan?"** This is one of the most common follow-up questions buyers ask assistants. If your page doesn't state it plainly, the model guesses, and it often guesses wrong in the direction that costs you the trial.
## A quick test before you touch anything
Open an incognito window, go to your pricing page, and select-all copy. Paste it into a plain text editor. What you see is roughly what a crawler sees. If the prices, plan names or limits are missing from that paste, they are missing for the models too, no matter how obvious they look on screen. This two-minute test has caught more invisible-pricing problems than any audit tool I know. Run it on your competitors' pages as well; you will often find theirs are perfectly legible to machines while yours is not, which is one reason assistants describe their pricing confidently and yours vaguely. If the copy test passes but assistants still misquote you, the problem is usually a conflicting third-party source outranking your own page, and the fix is updating that source rather than rewriting yours again.
## What a pricing page that teaches models well looks like
You don't need a redesign. You need the page to answer, in plain text, the questions buyers actually ask assistants:
- How much does it cost? State real numbers with the billing period attached.
- Who is each plan for? One sentence per tier, naming the buyer.
- What are the hard limits? Seats, usage caps, the features that differ.
- Is there a free plan or trial? Yes or no, with the terms.
- What happens when I outgrow a tier? Upgrade paths and overage behaviour.
Write these as short declarative sentences, not marketing copy. "The Team plan costs $49 per month and includes up to 10 seats" is quotable. "Built for ambitious teams ready to scale" is not.
A compact FAQ at the bottom of the pricing page does disproportionate work here. Each question a buyer would ask an assistant, answered in two or three plain sentences, is almost pre-formatted training material.
## Check what the models currently say
Before and after you rewrite anything, ask the assistants the questions yourself: "How much does [your product] cost?", "Does [your product] have a free plan?", "Is [your product] good for [your target customer]?" Note what they say and, where possible, what they cite.
Most teams are surprised twice: first by how much of the answer comes from the pricing page, and second by how often the quoted details are stale or wrong. If you want this monitored continuously rather than checked by hand, that's the problem I built Vouch for; it tracks what AI assistants say about your brand, pricing questions included. It's at vouchgeo.com.
## The compounding effect
A pricing page rewrite is a few hours of work. But unlike an ad campaign, it keeps paying: the updated page is read by every future crawl, quoted by every future answer, and absorbed into every future model refresh. In AI visibility terms, it is one of the few changes where a single afternoon of writing can correct what buyers hear for years.
Disclosure: I am the founder of Vouch (vouchgeo.com), a tool that monitors what AI assistants say about your brand.

