Your Pricing Page Is Invisible to AI (And It's Quietly Costing You Deals)

Quick Takeaways

• 51% of B2B software buyers now start research in an AI chatbot, and pricing is one of the first questions they ask it
• "Contact us for pricing" does not read as premium to an AI assistant. It reads as no data, so the assistant answers with a competitor's numbers or a guess
• JavaScript-rendered pricing tables and PDF rate cards are just as invisible: most AI crawlers read raw HTML, not what a browser renders
• The fix is unglamorous: a plain HTML table, a published starting price, a pricing FAQ, and the same numbers repeated consistently across your site
• Publishing a starting price does not kill enterprise deals. A "from $X/mo" anchor and a custom enterprise tier coexist fine, and Naoma's own pricing page is built exactly this way

You were not rejected. You were unreadable.

That is the uncomfortable explanation for a growing share of lost deals. 51% of B2B software buyers now start their research in an AI chatbot, up from 29% a year earlier, and one of the first things they ask is "how much does it cost?" If the assistant cannot read your pricing page, it does not say "pricing unavailable, go check the site." It answers anyway, with your competitor's numbers, an outdated figure from a third-party roundup, or a shrug that pushes the buyer toward vendors it can describe with confidence. In that same G2 research, 69% of buyers chose a different vendor than they originally planned based on AI guidance. Your pricing page has a new reader, and it does not fill out forms.

Pricing is the first question AI gets asked

Think about how a buyer actually uses ChatGPT or Perplexity in an evaluation. They rarely ask "tell me about Vendor X's brand values." They ask the questions a sales rep would normally field: what does it do, who is it for, what does it cost, how does it compare. We covered the shortlist mechanics in how B2B buyers use ChatGPT to shortlist vendors; pricing is the sharpest edge of that shift, because pricing questions have a right answer and buyers can tell when the assistant is guessing.

Price is also where shortlists get cut. A buyer comparing five tools will ask "which of these fits a $500/month budget?" and drop whichever vendors the assistant cannot place. The assistant is not being unfair. It is doing exactly what the buyer asked with the data it can see. The vendors with clean, published, machine-readable pricing get placed accurately. The rest get placed by rumor.

How AI assistants actually read a pricing page

There is no magic here, and understanding the mechanics makes the fixes obvious.

AI assistants get pricing information from two places. First, training data: what the model absorbed from the open web months ago, including old blog posts, forum threads, and review-site summaries about you. Second, retrieval: for current questions, engines like Perplexity, ChatGPT with browsing, and Google's AI Overviews fetch live pages and read them on the spot.

Both paths share a weakness: they overwhelmingly read raw HTML. When a crawler fetches your pricing page, it typically gets the document your server sends, not the page a human sees after JavaScript runs, cookies load, and a currency selector fires. If your pricing table is assembled client-side by a React component, hidden behind a region picker, or embedded in an image or PDF, the crawler often sees a headline, a hero paragraph, and an empty div where your prices should be.

So the assistant falls back on whatever else it has: a two-year-old Reddit thread ("I heard they start around $2k/month"), a review-site page with stale tiers, or a competitor's comparison post that frames your pricing however suits them. You do not control any of those sources. You only control your own page, and if that page is unreadable, you have delegated your pricing narrative to strangers.

"Contact us" no longer means premium. It means absent

For two decades, hiding pricing was a deliberate strategy: force the conversation, qualify the lead, negotiate from strength. Whatever you think of that logic with human buyers, it collapses completely with AI intermediaries.

When an assistant reads "Contact us for pricing," it records exactly one fact: this vendor publishes no pricing. Then the buyer asks the natural follow-up, "roughly what would it cost?", and the assistant does one of three things:

1. Estimates from third-party scraps. Old G2 answers, procurement forums, "what we paid" threads. These numbers are frequently wrong in both directions, and a too-high guess quietly disqualifies you from budgets you would happily serve.
2. Substitutes a competitor's published pricing as the category anchor. "Vendor A does not publish pricing, but comparable tools like Vendor B start at $249/month." Your competitor just became the reference point for your own category.
3. Recommends the vendors it can price. When the buyer asks "which of these fits my budget?", vendors with no readable pricing simply fall out of the comparison.

None of these outcomes involve you being evaluated and losing. They involve you never being evaluated at all. Buyers already prefer to self-serve this information: 67% of B2B buyers prefer a rep-free buying experience, and the AI assistant is the ultimate rep-free channel. Buyers doing their own qualifying is the core of the buyer-led sales motion, and hidden pricing opts you out of it. "Contact us" asks the one reader who cannot contact you to do the one thing it cannot do.

The three ways pricing pages go dark

Most invisible pricing pages fail in one of three ways, and plenty fail in two.

The gate. "Contact us," "Talk to sales," "Request a quote." No numbers anywhere. This is the most common failure and the most total: there is nothing to read, so the assistant reads someone else.

The JavaScript mirage. The page has real pricing, but it renders client-side: a framework hydrates the table, a toggle switches monthly and annual, a script localizes currency. Humans see a polished page. Crawlers see scaffolding. This one stings because the team believes the pricing is public. It is public to people, not to machines.

The PDF or image. Rate cards as downloadable PDFs, pricing screenshots in a slide, a table exported as a PNG. Some engines extract some text from some PDFs, but reliability is poor, and images are worse. If your only pricing artifact is a file, assume the assistant has not read it.

A quick self-test: open your pricing page, hit "view source" in the browser, and search for your own starting price. If the number is not in the raw HTML, an AI assistant probably is not seeing it either. Then ask ChatGPT and Perplexity "how much does [your product] cost?" and see what comes back. Many teams find this five-minute audit genuinely alarming.

What an AI-readable pricing page looks like

The good news: this is not exotic technical work. It is hygiene, and most of it ships in a day.

1. Put pricing in a plain HTML table. A real <table> element, server-rendered, with tier names, prices, and what each tier includes as text in the markup. Tables are the single easiest structure for engines to extract and quote accurately. Keep the interactive toggles if you like, but make sure a sensible default renders without JavaScript.

2. Publish a starting price, even if most deals are custom. "From $249/month" is enough to anchor you correctly in budget conversations. Without an anchor, the assistant invents one. With an anchor, you get placed in the shortlists you belong in, and enterprise buyers still call.

3. Explain the pricing model in one plain sentence. Per seat? Per usage? Flat tiers? Write it as a sentence a model can lift verbatim: "Pricing is pay per engaged demo, starting at $249 per month for 50 demos." Definitional sentences near the top of the page are what engines quote.

4. Add a pricing FAQ. Real questions buyers ask, answered in text: what counts as usage, what happens at the limit, is there an annual discount, what does enterprise add. FAQ blocks map directly onto the question-and-answer shape of chatbot conversations, which makes them disproportionately quotable.

5. Keep the numbers consistent everywhere. If your pricing page says $249, your comparison posts say $299, and your G2 profile says "contact us," the model sees conflicting evidence and hedges, or picks the wrong number. Audit every page and profile that mentions price, and update them together when pricing changes. Consistency is what turns a claim into a fact engines repeat with confidence.

A worked example: how we handle it at Naoma

We eat our own cooking here, so use Naoma's pricing page as a reference implementation, not because the prices matter to you, but because the structure does.

The model is stated in one extractable sentence: pay per engaged demo, billed only when a prospect engages for 3 or more minutes, with bounced visits free. The tiers are a plain table: Starter at $249/month for 50 engaged demos, Growth at $750/month for 100 demos with live product walkthroughs and meeting booking, Scale at $2,083/month billed annually for 300 demos and 2 agents, and a custom Enterprise tier with SSO, SLA, and security review. Edge cases that would otherwise become forum speculation are answered in text on the page: no surprise charges, demos pause at the limit. And there is an ROI calculator for buyers who want to model the math themselves.

Ask an AI assistant what Naoma costs and it has everything it needs to answer correctly: the model, the floor, the tiers, the caveats. That is the entire goal. Not to look cheap, not to give away negotiating leverage, but to make sure the answer a buyer hears is the one we actually wrote. The same logic applies one step later in the funnel, by the way: once an AI-referred buyer lands on your site pre-anchored on price, the next thing they want is to see the product, which is why we let them get an AI demo now instead of handing them a form.

"But we're enterprise": published floors and custom deals coexist

The standard objection: "Our deals are six figures and every contract is custom. We can't put a price on a page."

You do not have to publish your rate card. You have to publish an anchor. "Plans start at $X/month; enterprise pricing is custom" gives the assistant something true to say and still leaves every negotiation open. Naoma does exactly this: three published tiers and a custom Enterprise plan sit on the same page without tension. The published floor does the AI-facing work; the enterprise tier does the deal-shaped work.

The honest trade-off is real but smaller than feared. Yes, a published floor lets some prospects self-disqualify. Most of those prospects were never going to buy, and they currently cost you sales time to disqualify by hand. What you gain is being quotable at the exact moment a buyer with real budget asks an assistant to build a shortlist. In a world where the first pricing conversation happens between your buyer and a model, being absent from that conversation is the expensive option.

The takeaway

Your pricing page now has two audiences: the buyer, and the AI assistant briefing the buyer before you ever know they exist. The second audience cannot click a currency toggle, cannot open a PDF reliably, and will never contact sales. Serve it with embarrassingly simple tools: a server-rendered HTML table, a published starting price, a one-sentence description of the model, a pricing FAQ, and consistent numbers everywhere your price appears. Then test yourself the way a buyer would: ask the assistants what you cost, and fix whatever they get wrong. The vendors winning AI-era shortlists are not the ones with the cleverest pricing strategy. They are the ones whose pricing an AI can actually read.

FAQ

Why can't AI assistants read my pricing page? Most AI crawlers read the raw HTML your server sends, not the page a browser renders. If your pricing loads via JavaScript, sits behind a "contact us" gate, or lives in a PDF or image, the assistant sees no numbers and answers pricing questions from third-party sources or competitor pages instead.

Should we publish pricing if all our deals are custom? Publish a starting price, not your full rate card. "From $X/month, enterprise custom" anchors you correctly in AI-built shortlists while leaving negotiations untouched. With no anchor, assistants either guess your price or use a competitor's published pricing as the category reference.

How do I check whether AI can read my pricing? Two tests. First, open your pricing page, view the raw source, and search for your starting price; if it is not in the HTML, crawlers likely cannot see it. Second, ask ChatGPT and Perplexity "how much does [your product] cost?" and compare their answers to reality.

Want the buyers your pricing page wins to see the product the moment they land? Get an AI demo now →