A buyer on your shortlist types three questions into an assistant before they ever fill in your form. What does this cost? Who does this well for a company our size? Is this vendor a fit for us? The assistant answers all three in under a minute. If your site stays quiet, the answer still comes, built from someone else’s page.
Half of B2B software buyers now start research in an AI chatbot more often than in Google, up from 29% a year earlier (G2, April 2026). 1 Source 1 G2, “New G2 Research: Half of B2B Software Buyers Now Start Their Research With AI Chatbots,” April 15, 2026. 1,076 B2B decision-makers surveyed in March 2026. 51% start with chatbots more often than Google, up from 29% in April 2025. And by the time they talk to you, most have already decided. 6sense found 94% of buying groups ranked their shortlist before contacting a seller, and the favorite going in won 77% of the time. 2 Source 2 6sense, “The B2B Buyer Experience Report for 2025,” November 2025. Nearly 4,000 responses. First contact at 61% of the journey; 94% ranked the shortlist before contact; the pre-contact favorite won 77% of the time.
So the money question gets answered early and without you. The only thing left to decide is whose answer it is.
The AEO playbook skipped the pages that close deals
Nearly every piece of AEO advice is about articles. How to structure a blog post, how to write a definition, how to build a topic cluster. We have written plenty of it ourselves.
The pages that decide revenue get left out. Service pages, pricing pages, comparison pages, and the FAQ that handles commercial objections. 3 Definition 3 Money pages: the pages a buyer reads when deciding whether to buy, as opposed to learning about a topic. Service, pricing, comparison, and the FAQ that answers commercial questions. They get treated as brand territory, owned by whoever owns the positioning, and they are usually the least readable pages on the site for a machine. Articles answer “what is.” Money pages are supposed to answer “how much,” “for whom,” and “why you and not them.” Those are exactly the questions a buyer asks an assistant in the last stretch before a shortlist locks.
For a CMO this is a pipeline problem. For RevOps it is a data problem, because the qualification that used to happen on the first call now happens in a chat window you cannot see, working from inputs you did not supply.
What does an AI engine do when your price isn’t on the page?
It answers anyway. It does not return a blank, and it rarely says “this vendor does not publish pricing, please contact them.”
It goes looking for the next-best source. On vendor-discovery questions, DerivateX found ChatGPT cited the recommended vendor’s own website 11.6% of the time. The other 87.4% of citations credited a third party: review sites, listicles, comparison posts, somebody’s roundup from last spring. 4 Source 4 DerivateX, “B2B SaaS AI Citation Study,” May 31, 2026. 40 categories, 169 cited URLs. A separate July 2026 study by Growfusely found vendors’ own pages made up 24% of ChatGPT citations and 12% of Perplexity citations across 128 product queries. Both are agency studies with modest samples. When the vendor page has nothing quotable on price, those third parties are the whole answer. If none of them has a number either, the engine reaches for whatever it absorbed in training, which may be a price you retired two years ago.
Buyers notice. G2 found 64% of B2B software buyers run into AI inaccuracies weekly or more. 5 Source 5 G2, “Your Buyers Are Using AI to Find Software. Here’s What They’re Trusting,” June 26, 2026. They also do not assume the best. TrustRadius has found transparent pricing to be buyers’ top request of vendors four years running, and its 2026 report notes that when pricing is opaque, buyers tend to assume it is high or negotiable, which is how vendors fall off shortlists. 6 Source 6 TrustRadius 2026 B2B Buying Disconnect Report, released July 15, 2026. 1,862 buyers and 444 vendors. The 45% figure for 2026 is from Agile Brand Guide’s summary, July 20, 2026.
Put those together and “contact us for pricing” stops being a neutral choice. It used to mean the buyer had to call you. Now it means the buyer asks a machine, and the machine asks someone else.
Contact-us pricing was a lead-capture device
We spent 16 years inside the inbound playbook, and it had a clear view of the pricing page: hide the number, make the buyer raise a hand, and route the hand to sales. The gated price became a form fill, and the form fill became a lead on a dashboard somebody was bonused against.
That logic assumed the buyer had nowhere else to go. They do now, and they would rather not call. Gartner found 61% of B2B buyers prefer a buying experience with no sales rep at all. 7 Source 7 Gartner, “Gartner Sales Survey Finds 61% of B2B Buyers Prefer a Rep-Free Buying Experience,” June 25, 2025. 632 buyers surveyed August to September 2024. A hidden price no longer produces a conversation. It produces a question typed somewhere else, and an answer you had no hand in.
”Contact us for pricing” used to mean the buyer had to call you. Now it means the buyer asks a machine, and the machine asks someone else.
A price the engine can’t see is an unpublished price
Plenty of companies do publish pricing and still lose the answer, because of how the page is built.
The common version is the interactive pricing page. A monthly and annual toggle. A slider for seats or page count. A tab per plan. A calculator that produces a number once you answer four questions. All of it runs in JavaScript, and Vercel’s analysis of AI crawler traffic found none of the major AI crawlers execute JavaScript, including OpenAI’s, Anthropic’s and Perplexity’s. 8 Source 8 Vercel, “The rise of the AI crawler,” December 17, 2024. The crawlers fetched JavaScript files but did not run them. The assistant reads the raw source of the page and never runs the scripts. If the number arrives after the script does, it is not in the answer. The buyer sees a clean price table and the engine sees an empty container.
The second version is subtler. Someone reads that structured data helps AI, so the price goes into markup behind the page while the visible page still says “let’s talk.” That does not work either. In OtterlyAI’s March 2026 test, a fact placed only in structured data was used by zero of seven AI platforms. 9 Source 9 OtterlyAI, schema markup experiment, March 23, 2026. Unique-information test: 0 of 7 platforms used data that existed only in structured markup. Markup can describe a price the reader can see. It cannot replace one.
The fix for both is architectural, and it is the same fix content-first rendering makes for every other page: the words, including the numbers, ship in the HTML the server sends. The calculator can still exist. It just cannot be the only place the price lives.
Publish the range, not the rate card
Nobody is asking a services firm to post a menu. Most B2B work is scoped, and a single fixed number would be wrong for half your buyers.
What an engine needs, and what a buyer is actually asking, is narrower than that. Roughly what does this cost, what drives it up, and what is included. That is answerable on almost any service page:
- A starting point or a typical range, stated as a sentence (“Most projects land between $X and $Y”)
- The two or three things that move the price, named plainly
- What the fee includes and what it does not
- Who it is not for, which is the fit answer buyers most want and vendors least often give
This is answer-first writing applied to money. Each of those lines survives being lifted out of the page on its own, which is what a citation engine is shopping for. 10 See also 10 The passage-level mechanics are in How LLMs Decide Which Sources to Cite. A range also filters. The buyer whose budget is a third of your floor finds out from the assistant, not from your AE forty minutes into a discovery call.
Does publishing a price hand your competitors an advantage?
Less than keeping it private does. Your competitors can already get your number from one sales call or one former client. The only party your silence reliably keeps in the dark is the assistant, and the assistant is the one talking to your buyer.
There is a real case for scoping every deal by hand, and in enterprise procurement with custom terms it holds. Even there, the page can say what the engagement shape is, what it typically costs to start, and what triggers a bigger number. “Engagements start at $X and scale with Y” gives nothing away that a competent competitor does not already know.
Service and comparison pages answer “who” and “which”
The price question gets the attention, but the fit question costs more deals. “Who does this for a company like ours” gets answered from service pages, and most service pages are written to be felt rather than parsed. “We partner with ambitious brands to unlock growth” gives an engine nothing to cite. It cannot tell who you serve or what they get.
A service page an assistant can use names the buyer (industry, size, situation), the problem, the deliverable, and the timeline, each in a sentence that stands on its own. Comparison pages work the same way. The version that gets used is the one that says plainly when a competitor is the better choice, because that is the passage that reads as a fair answer rather than an advert. The FAQ belongs in visible text on the page, where both buyers and machines can read it.
What we did on our own pages
Our AI Strategy page publishes its prices in the page source: a $2,500 fixed-fee readiness audit and a $15,000 to $25,000 strategy engagement, with what each includes. 11 Editor's note 11 We checked this the same way we suggest you check yours: fetching the page as a crawler would, without running scripts, and looking for the number in what came back. Both numbers are in the HTML a crawler receives, not assembled by a script.
We have not done this everywhere. Our Sovereign Stack page answers “how does pricing work?” with a paragraph about project fees and scoping and no number. By the standard in this piece, that page is handing the answer to someone else, and it is next on our list.
What you gain when the money pages answer
We will not promise you a conversion rate, because nobody honestly can. What changes is who controls the answer.
When your pages carry the range and the fit criteria in plain text, the assistant has something of yours to quote besides a two-year-old roundup. The number a buyer arrives with is your number. The buyers who were never going to fit remove themselves before they cost your team a discovery call, and the ones who stay start the conversation further along. For RevOps, the first touch gets cleaner. The price objection that used to surface in week three has already been settled, one way or the other, by something you wrote.
That is the real asset: the buyer’s money question answered on your terms, by your page, before the shortlist locks. The search for a conversion lift can come later. The page a buyer needs has to exist first.
Want to see what an assistant can actually read on your pricing and service pages? The Free AI Website Assessment checks in 60 seconds.
Start with a test that takes ten minutes. Ask two assistants what your core service costs, who you are a good fit for, and how you compare to the competitor you lose to most. Write down whose pages they cite. Then open your pricing page, view the source, and search for your own number. If it is not there, neither are you.
For the complete build in order, How to Get Found by AI is the guide this piece sits under.
The research is free. So is the diagnosis. 60 seconds →