How to make your SaaS pricing page readable by AI assistants

Buyers now ask an AI assistant what you cost before they open your site. If the assistant can't read your pricing page, it answers from a forum, a review site or a price tracker instead. Run the Agent Price Read: ask the assistants, read the page raw, open the door to agents, put prices at the top, and fix the outside sources.

Key points

• A pricing page that looks perfect in your browser can be empty to an agent that reads raw HTML.

• In one 2026 study, company pricing pages showed up in 46% of AI pricing answers but came first in only 12%.

• Agents run on a clock and a bill. Put plan names and prices in plain text near the top of the page.

• Fixing your own page is step four of five. The outside sources still get quoted, so fix them too.

Why does an AI-readable pricing page matter now?

Because the first price your buyer sees may not come from you. Pricing experts and AI search practitioners spent 2026 measuring this, and the numbers are hard to ignore.

Growth Unhinged is a newsletter on SaaS growth and pricing that Kyle Poyar writes for founders and operators. In September 2026, Kyle and Nikolas Laskaris, a researcher at Profound, published a study of 7,600 AI responses to pricing questions about the Cloud 100, a list of top private cloud companies. Profound is an AI marketing platform that tracks how brands show up in AI answers for marketing teams.

Here is what they found:

• Six engines, two weeks. The data covered six AI engines, including ChatGPT, Perplexity and three Google surfaces, collected 29 July to 10 August 2026.

• A 24x spread on one product. For one company with no public pricing page, AI answers ranged from roughly $100 to $2,400 per user per month.

• Your page shows up, but rarely first. Company pricing pages appeared somewhere in 46% of responses and first in only 12%.

• Strangers fill the gap. The most cited outside sources were a price-tracking site at 18.7%, a forum at 18.6% and a review site at 15.9%.

ChatGPT is an AI assistant that answers questions in chat for consumers and businesses. Perplexity is an AI answer engine that searches the web and cites its sources for researchers and buyers. Google is a search engine that now puts AI-written answers above its links for everyone who searches.

One caveat, the same one I give in the original post: a co-author works at a company that sells AI visibility tooling, and Profound published the same research on its own blog. Kyle and Trevor Pyle of Profound also ran a webinar on it on 8 October 2026. Treat it as a strong signal, not settled science.

A second benchmark shows the mechanism. Search Engine Journal is a publication that covers search marketing news for SEO and digital marketers. Its June 2026 write-up of a vendor's report found that 65% of plans showed readable prices, about 30% of runs hit at least one error, and runs with access errors pulled 58% of their content from third parties, against 12% for clean runs. The median run took about 32 seconds and cost $0.24.

An agent isn't going to dig around your site. It has a clock and a bill running. What it costs you is anchoring: your buyer arrives on the first call already holding a number you didn't write.

What is the Agent Price Read?

As I put it in the original Agent Price Read post: "The framework: the Agent Price Read, 5 steps you can run in an afternoon." You need an afternoon, a browser and whoever owns the page and the site's bot settings.

The goal is a pricing answer that humans and agents both read the same way.

How do you run the Agent Price Read, step by step?

Work through the five steps in order. Step 1 gives you the before shot, and every later step gets measured against it.

Step 1: Ask it

Open 2 or 3 AI assistants and ask the question a buyer would ask: "How much does [your product] cost for a 50 person team?" Write down exactly what each one says and which sources it points to.

Do this before you touch anything. It tells you whether you have a problem and gives you a before shot to compare against later.

Run the question three times on each assistant. Answers vary from one run to the next, so three runs give you a better read than one.

Step 2: Read it raw

Check what the page looks like without JavaScript. Open the page source, or turn JavaScript off in your browser, and search for your actual prices. If they're gone, an agent that reads raw HTML never saw them.

The study's authors point at server-side rendering as the fix, and the benchmark's advice is the same. Every section with text on it should render on the server. That's an engineering ticket, not a copy change, which is why you want to find out now.

Step 3: Open the door

Look at who is allowed in. Check your robots.txt and the default bot settings on your CDN or firewall for rules that block unknown or automated visitors from the pricing URL.

Rules written to stop scrapers can catch assistants too. A locked door doesn't just hide your page. It sends the agent to someone else's.

Don't open everything to everyone just to pass this step. Allow the pricing URL and the pages that explain it, and leave the rest of your rules alone. My guide on how to get cited by ChatGPT and Perplexity covers crawler access across the rest of the site.

Step 4: Front-load it

Put plan names and prices in plain text near the top of the page. The benchmark says agents typically pull only the first 15,000 to 20,000 tokens, so a price below a long hero, a logo strip and a feature grid may never be read.

Then build what the study's authors call a pricing answer stack:

1. The pricing page, with prices in plain text.

2. A pricing FAQ that answers seats, billing and plan limits.

3. Billing docs that say the same thing in plain language.

An agent that finds the same answer in three places stops looking for a fourth.

Step 5: Fix the sources

Read what outside sources say about your price. Search your product name plus "pricing" and open the review site, the price-tracking site and any forum thread that comes up.

When one is wrong, correct it where you can, or answer it publicly and plainly. Those pages are most of what an agent quotes when yours fails.

Put the whole read on a monthly reminder. Pages change, bot rules change and assistants change, so re-run steps 1 to 5 each month against your before shot.

This is the part most teams skip, and it's the part I build into client work. Content Engineering is my service that gets your brand cited in the sources AI search and buyers read, for founders and operators at SaaS companies. Kevin Payne is an AI engineer and GTM operator who installs AI workflows for B2B SaaS revenue teams. In Phase 1 Strategy we agree the buying questions you want to show up for, including "what does it cost," check which pages AI engines cite today, and plan fixes for wrong details on the outside pages they quote. Book a free coffee chat to get yours scoped.

What are the most common mistakes?

Most mistakes come from trusting what you see in your own browser. Here are the ones the original post calls out, with the fix for each.

• Asking once and moving on. One run is noise. Fix: three runs per assistant, logged with sources.

• Trusting how the page looks. It looks perfect in your browser, and that's the whole problem. Fix: view source and search for the price.

• Opening every door. Loosening all bot rules to pass step 3 creates new risk. Fix: allow the pricing URL and the pages that explain it, nothing more.

• "Contact sales" on every plan. Agents fill the blank with someone else's number. Fix: if some tiers are custom, say so in plain text and publish the ones that aren't.

• Fixing your page and stopping. You fixed the first source. The assistant still has two others to quote. Fix: run step 5 and the monthly re-run.

What does this look like in practice?

It looks like one afternoon, one sheet and three owners. When I scope this work, I start with step 1, because the before shot decides whether the rest is urgent.

Copy this sheet into a doc and fill it in:

• Question asked and assistants used:

• What each one said about price:

• Sources each one cited:

• Prices visible in raw HTML (yes or no):

• Pricing URL allowed in robots.txt and CDN settings (yes or no):

• Plan names and prices in the first screen of plain text (yes or no):

• Pricing page, FAQ and billing docs match (yes or no):

• Wrong or stale numbers found on outside sources, and who will correct them:

• Date of next re-run:

Send the filled sheet to whoever owns the page, the bot settings and the billing docs. If you're deciding who should own this long term, my comparison of content engineer vs content marketer walks through which role fixes outside visibility. For the wider discipline, start with what content engineering is.

When does the Agent Price Read not apply?

It applies less when you never publish a number or don't control your listing. Here is how to adjust.

• Fully custom pricing. Steps 2 to 4 won't help much. Focus on step 5 and publish whatever ranges and principles you're comfortable with.

• Partner or marketplace listings only. Run steps 1 and 5 and hand the rest to whoever controls the listing.

• A few named inbound accounts. This is lower priority than for a team with broad top-of-funnel demand. Do it anyway, behind real fires.

The research is also young. It comes from two sources, both run by vendors in this space, and assistant behavior will keep shifting. That's another reason for the monthly re-run.

FAQ

How do I check what ChatGPT says my SaaS costs? Ask it the way a buyer would, for example "How much does [product] cost for a 50 person team?" Run it three times, note each price and every cited source, then repeat in Perplexity and Google. That log is your before shot for every later fix.

Why can't AI assistants see the prices on my pricing page? Often the prices load with JavaScript after the page renders, so an agent reading raw HTML never sees them. Bot rules in robots.txt or your CDN can also block it. View the page source and search for your prices to find out.

Should I publish prices if my plans are custom? Publish what you can. If some tiers are custom, say that in plain text and publish the ones that aren't. A blank "contact sales" on every plan leaves the agent to quote a forum, a review site or a price tracker instead.

How often should I re-run the Agent Price Read? Monthly. Pricing pages change, bot settings change and assistants change how they answer. Re-run all five steps on a calendar reminder and compare each result against the before shot you took in step 1, so you catch drift before a buyer does.

Get your pricing answer checked

If you want help finding where AI assistants get your price wrong and fixing the pages they quote, book a free coffee chat. You can also see every way I work on the services page.

About the author: Kevin Payne is an AI engineer and GTM operator who installs AI workflows for B2B teams. He has 10+ years in B2B SaaS demand generation and RevOps across Techstars, a16z and YC portfolio companies, and his writing has appeared in HubSpot, GetResponse, Outbrain and more.