We reviewed 10,777 business websites and found a pricing gap
Only about 5% had prices labelled for software to read. Here is what that means for your business, how to make your offers easier for AI assistants to recommend, and where Nexez can help.
Reviewed October 7, 2026 · 11 min read
On this page7 sections
A potential customer asks an AI assistant: “Find a house cleaner near me, tell me what it costs, and help me book.”
Your website might look great. It might have glowing testimonials and a prominent “Contact us” button. But can someone answer that request from what you have actually published? What does the cleaning include? Which neighborhoods do you cover? What changes the price? Is the next step a confirmed booking or just a request?
Those are sales questions, not technology questions. They are also the questions your website needs to answer when software helps a customer shop.
Our latest study covers 10,777 unique business websites that passed our content review. Across three groups, our checks found a visible price on 21.7% to 23.0% of the pages examined. Only about 5% had price information in the labelled format our scanner looked for.
The opportunity is not to add more AI buzzwords to your website. It is to make the offer clearer.
The headline numbers#
The same broad pattern appeared in each group:
- A visible price appeared on 21.7% to 23.0% of pages.
- Pricing labelled for software appeared on 4.9% to 5.2%.
- A booking, ordering or other next-step link or form appeared on 60.6% to 63.2%.
- An action described using labels software could recognize appeared on 34.5% to 35.3%.
- Readable behind-the-scenes website data appeared on 68.6% to 69.3%. That did not necessarily include useful details about what a customer could buy.
- A way to contact the business appeared on 80.1% to 81.1%.
- A stated policy, such as returns or cancellation, appeared on 7.6% to 9.3%.
These figures come from the original September 30 to October 4 observations for websites that passed our October 5 content review. The ranges show three separately reported groups, not a national average.
Put two of those lines side by side and the gap is hard to miss. These businesses were roughly sixteen times more likely to publish a way to contact them than to publish what they charge in a format software can read. Contact details are table stakes. The price is the thing the customer actually asked for.
The policy line is the lowest of the set. Cancellation terms, return windows and similar rules are exactly what a careful buyer compares before committing, and fewer than one page in ten stated one in a form our checks could detect.
The practical takeaway: publishing a website, explaining an offer and making that offer easy to act on are three different jobs.
The result that was missing entirely#
Most of the findings above are about information that is present but hard for software to read. One finding is different, because the thing we looked for was simply not there.
Alongside ordinary page checks, our scanner looked at the conventional locations for agent manifest files. These are small machine-readable documents that describe what a business offers and how software can act on it. Across all 10,777 websites, it found none. No root agent.json, no well-known agent.json, no agent card. Two related checks for shapes resembling an MCP server or an OpenAPI description returned three hits each, out of 10,777.
What this does and does not show
These were shallow checks at fixed, conventional paths. A business could run an integration somewhere our scanner did not look, and a missing file is not a judgement about the business. What the result does show is that at the locations software would conventionally check first, this sample published nothing. That is a different problem from a badly written page, and it has a different fix.
It is worth being plain about why this matters and why it might not. These files are not a requirement for appearing in search or in an AI answer, and no business is being penalised for lacking one. But they are the part of the stack designed to let software transact rather than merely read, and in this sample that part was empty.
What an unclear website can cost you#
Imagine two cleaning companies.
One says “Quality cleaning at affordable prices. Contact us.”
The other explains its service area, what a standard clean includes, its starting price, the factors that change the total, and how to request a slot.
The second company gives a customer more to compare. It also gives an assistant more evidence for answering a specific request. That does not guarantee a recommendation, but it removes avoidable uncertainty.
When those details are missing, the risks are familiar:
You are harder to compare. A budget-conscious customer cannot tell whether you fit without starting a conversation.
You attract avoidable enquiries. People contact you about locations, services or budgets you do not support because your site never rules them out.
The handoff breaks. Someone finds the right service, but the button sends them to a generic homepage, an old booking form or a page that explains nothing about what happens next.
Expectations drift. An outdated price or unclear cancellation policy can turn an interested buyer into a disappointed one.
These are practical risks, not lost-sales figures measured by our study. They are also problems worth fixing whether the visitor is a person or an AI assistant.
Find the gaps on your own website
The free Nexez scan checks accessible page information, including pricing, business details and next-step signals. Use it as a starting point for improvements, not a promise of where an AI system will rank you. No signup required.
Scan your website freeHow to make your business easier for AI assistants to recommend#
You cannot control an assistant’s shortlist. You can control how clearly and consistently your business presents the facts a customer needs.
Start with these six improvements.
1. Give each important offer a clear page#
Do not make a customer assemble your offer from a homepage, a social post and an old brochure.
For each main service or product, explain what it is, who it is for, what is included, where it is available and how to take the next step. Answer real buying questions, such as “Do you bring supplies?” or “Can I cancel after ordering?”
Use a descriptive page title, a clear main heading and ordinary text for essential details. Link to the page from your navigation or relevant pages. These are useful search basics, not a special AI trick. Google’s SEO guide recommends clear, helpful content and descriptive links.
2. Explain the price even when you cannot give a fixed quote#
Not every job has a flat price. You do not need to invent one.
Publish a genuine starting price, an honest range, a package price or a clear explanation of how quotes work. Say which details change the amount and what the customer needs to provide.
For example, “Move-out cleaning priced by property size, with oven cleaning quoted separately” is more useful than “Competitive rates.”
Where you show a starting price, make the conditions obvious. The goal is a better-informed enquiry, not a misleadingly cheap headline.
3. Add the right labels behind the scenes#
The technical name is schema markup. Think of it as labels that explain your page: this is the business name, this is the service, this is the price, and this is the purchase link.
Ask your website provider for the types that fit your business:
| Your website describes | Markup to ask about |
|---|---|
| A local business and its location or hours | LocalBusiness |
| A service and the offer attached to it | Service and Offer |
| A product, price and stock information | Product and Offer |
You do not need to learn the code to ask the right question. Ask: “Does our website label the things we actually sell, or does it only label our company name?”
That question matters because of one pairing in our results. Readable behind-the-scenes data appeared on roughly 69% of pages, while pricing labelled for software appeared on about 5%. Most of these sites already had labels. They described the company rather than the thing being sold.
Google documents local business information and product information; Schema.org describes services and their offers.
Keep the labels consistent with the page. Do not invent prices, availability or reviews to fill empty fields. Product markup can make an eligible page suitable for richer Google results, but adding labels is not a ranking guarantee or proof that an assistant can buy.
4. Make your public information accessible and consistent#
Check that your important pages are not accidentally hidden from search. Keep essential offer details available as text and keep your business profiles current.
Google says its AI search features use the same core SEO practices as ordinary search. There is no special AI markup or extra text file required to appear in them. Read Google’s guidance.
Have your website provider review access settings for the search services you want to reach. Do not disable security or open private customer information just to improve a scan score.
5. Make the next step match the offer#
A customer reading about a specific service should reach that service’s booking, purchase or quote process.
Explain what happens after they click. Are they paying a deposit, requesting availability or confirming an appointment? What information will they need? When will they hear back?
Test the process on your phone. A working homepage does not tell you whether the booking page still works.
6. Check the result and keep it current#
Run another scan after changes. Ask your developer to check supported markup with Google’s Rich Results Test. That test checks eligible search features, not every possible use of your data.
Try a few realistic customer questions in the assistants your customers use. Inspect the answers and cited pages. Treat this as a diagnostic exercise, not a reliable ranking test: one answer is not a permanent position.
Then measure what matters to your business: relevant enquiries, completed bookings and purchases. Update prices, hours and policies when they change, not just when you redesign the site.
How Nexez helps close the gap#
You do not have to replace your website to make your offers clearer.
Nexez helps turn what you already sell into listings with explicit services, prices, options and next steps. You can start with your existing website or enter the information yourself, then review the draft before publishing. Start with your website.
It organizes the offer. Instead of leaving a buyer to piece together scattered pages, your listing brings the service, description and choices together.
It generates the labels. Nexez creates structured markup from listing information, including supported offer, price and action details. That applies to the Nexez listing; it does not silently repair every page on your existing website.
It puts your rules around the sale. You set the prices, options and requirements. Configured offers can collect the buyer’s answers, check whether the request fits and provide the supported next step. Checkout depends on the offer and required setup being ready. See how Nexez works.
It lets you test before relying on an order. The simulator helps reveal missing or unclear information. It is a test environment, not proof that an outside assistant has recommended you or completed a real transaction.
For a cleaning business, the work is concrete: define the service, explain which property details affect the price, set the work you accept, and give the buyer a clear route forward. Nexez helps you express those decisions. It does not make the business decisions for you.
Try the Nexez simulator or create your business listing.
Being recommended and being ready to sell are different goals#
A search engine can link to your business without understanding every price. An assistant can recommend a service without being able to book it. A booking link can exist without a reliable checkout behind it.
That is why “agent ready” should not be treated as a magic badge.
Our recommendation is to work through the practical steps in order: make your business findable, explain the offer, make the facts consistent, then test the next action. Nexez supports that work. It cannot promise inclusion in every AI answer, a particular ranking or a sales increase.
The study’s value is a clearer set of priorities. Before buying another AI marketing tactic, check whether your own website answers the basic questions a customer is already asking.
Scan your site free · See how Nexez helps · Get listed
About the study#
The study includes 10,777 unique business websites that passed our October 5, 2026 content review, with earlier observations collected September 30 to October 4. “Unique” refers to website domains, not separately verified companies. We did not certify active trading, ownership or industry labels. Results are reported across three groups because original selection rules differed; this is not a national estimate or a trend comparison with the earlier 652-site study. We checked pages and signals, not completed purchases.
One detail about those three groups is worth stating. They were selected under different rules, observed on different dates, and ranged in size from 1,350 to 4,724 websites. Even so, the labelled-pricing figure landed at 4.95%, 5.17% and 4.89%, and the mean readiness scores at 52.44, 52.25 and 52.19. We do not combine the groups into a single national figure, and nothing here supports one. But three separately drawn groups agreeing that closely is a reason to think the pattern reflects how business websites are commonly built, rather than an artifact of how any one group was chosen.
The full accounting, exact percentages and limitations remain available in the analysis and methodology. The aggregate figures and review counts are also published as machine-readable files alongside those pages, and collection for this round is closed, so the numbers above can be checked against the same fixed set we used.
Make your best offer easier to find and buy
Start with one offer, not a complete website rebuild. Check it, clarify it, and give the buyer a next step that works.
See how Nexez helpsFrequently asked questions
Will schema markup make AI assistants recommend my business?
No markup guarantees a recommendation. It can describe your content more explicitly for systems that use it. Accurate, helpful public information and a usable website still matter. Google does not require special schema for its AI search features.
Do I need to publish every price?
No. If the amount depends on the job, explain the starting point, what changes the price and how to get a quote. Never publish a fixed amount you cannot honor.
Does Nexez replace my website?
No. It adds a clearer listing and supported buying paths alongside your existing site. Changes to your original site’s content or markup still need to be made there.
Is the free scan a test of every AI assistant?
No. It checks defined website signals. It does not reproduce every assistant’s search, browsing or purchasing behavior, and it does not complete a purchase.
What does the study count as a unique business website?
It counts separate website domains that passed our content review, not independently verified companies. One business can own several sites, and passing the review does not certify active trading, ownership or industry. The supporting methodology explains the full accounting.
Keep reading
How to stop AI assistants booking jobs you don’t take
An assistant can only follow rules you have written down. Say what work you take, what you need to know and what needs your OK.
7 min readGuidesWhat it costs a small business to ignore AI assistants
The real cost of ignoring AI assistants is the customers you never hear from. How to work it out, and what acting costs instead.
8 min readGuidesQuotes and estimates: how to price custom jobs so AI assistants can share them
How to replace “call for a quote” with estimates AI assistants can share, and when a job still needs a firm quote.
9 min read