We read 828 AI answers to questions about cost
Should you publish your prices so AI engines can quote them? Plenty of advice says an engine won't recommend a company that hides its pricing. We found little measurement behind it, so we looked at what the engines say when a buyer asks what something costs.
The sample is buyer questions our users track that ask a price outright: "How much does a 24x7 SOC cost per month?", "What is the price of an IPL laser machine?", "and how much do they cost?". That came to 101 questions for 56 brands, asked on ChatGPT, Claude, Perplexity and Google AI Overviews from October 4 to November 6, 2026. After dropping repeated identical answers, 828 remained. We compared them with 10,089 answers to the same brands' other buyer questions.
99.6% of the answers quoted a price. 49.8% named no company.
825 of the 828 answers stated a figure, in a currency or per month or per hour. On the other buyer questions, 29.9% did.
412 answers (49.8%, 46.4% to 53.2%) named no company, product or provider at all. On the same brands' other buyer questions, 8.8% named nobody. The median cost answer named one company; the median other answer named five.
A typical one: asked what an outsourced security operations centre costs a growing business in India, ChatGPT gave four package tiers with monthly rupee ranges and what each covers. No provider appeared. The buyer got a budget, not a shortlist.
The engines differed. Perplexity named nobody in 66.2% of its cost answers, Claude in 53.6%, ChatGPT in 45.6% and AI Overviews in 31.1%.
88% of the nameless answers cited a page about cost
The numbers came from somewhere. 364 of the 412 answers that named nobody cited at least one page whose address was about cost, price, fees or rates. Across all 828 answers, 3,397 of 9,113 citations were pages like that: 37%.
Reading the answers by hand, several of those cost guides sat on the blogs of agencies, installers and marketplaces selling the very service in question. The engine took their number and left their name behind.
When a company was named, its own pricing page was cited 20% of the time
The cost answers named 890 companies we could match to a website. For 180 of them (20.2%, 17.7% to 23.0%) the same answer cited a pricing, plans or cost page on that company's own site. On the other buyer questions the figure was 632 of 25,446 named companies: 2.5%.
Put the other way, 108 of the 416 cost answers that named anyone cited a named company's own price page: 26%. When an engine names a supplier while answering a price question, it often has that supplier's price in front of it.
The tracked brands were named in 4.1% of cost answers
The 56 brands themselves were named in 34 of the 828 cost answers: 4.1% (3.0% to 5.7%), against 10.7% on their other buyer questions. 35 brands did worse on cost questions, 6 did better and 15 were never named on either.
14 of the 101 cost questions never produced a single company name on any engine over the month.
A study of branded pricing questions found the same pull toward outside pages
In August 2026, Growth Unhinged and a research partner asked six engines to "Evaluate Company X on pricing" for the Cloud 100 and read 7,675 answers. The company's own pricing page appeared in 46% of them but came first in only 12%. Vendr, Reddit and G2 were each cited in about 16% to 19%.
That study covers buyers who already know the name. Ours covers the question before it, when the buyer is still asking what the thing should cost.
What we'd do
- Publish a price, or a starting price and what moves it, as plain text on a page crawlers can fetch. Named companies had that page cited eight times as often on cost questions as on other questions.
- Write the cost guide for your category: the ranges, what drives them, and where you sit. Engines quote these pages. Our data can't show they'll keep your name, only that they read the page.
- Read your cost answers, not just the rate. An answer that names nobody can still be quoting your page.
How Proofsource handles it
Proofsource keeps every answer and every URL it cited, by engine and question, so you can see whether a cost answer used your page, and whether it said your name.
Where this could be wrong
We found cost questions by phrase, so a question that asks about price in other words was missed. The 101 questions lean toward services, many in India, from brands our users chose to track.
Names come from our extraction, which can miss one; in a hand read of 14 nameless answers we found none missed. A pricing page is matched by words in its address, which is crude.
This is correlation over one month and four engines. Companies with public prices may differ in other ways, and we can't show that publishing a price causes an engine to name you.
