What does "agent-ready" mean for a small business? When ChatGPT, Gemini or Yelp Assistant asks about price and availability

For a small business, "agent-ready" means one thing you can test: when an AI assistant asks on a customer's behalf what you charge and when you are free, it gets an answer. That answer needs a price or range, availability, the policies that matter, your service area, any minimums and a real next step, in writing and within minutes. The measured baseline is poor. When Google's AI phoned businesses for prices in late 2025, 26% of calls went unanswered and 48% of answered calls gave no price.¹ On our arithmetic, about 38% of those requests ended with a price.

What is being misread about "agent-ready"?#

The phrase "agent-ready" is mostly sold as a technical project: structured data, product feeds, an llms.txt file. The evidence points elsewhere. Google's guidance for AI Overviews and AI Mode says: "You don't need to create new machine readable files, AI text files, or markup to appear in these features."⁴ Ahrefs studied 137,210 domains in May 2026 and found 28% published an llms.txt file, of which about 97% received no requests at all; AI retrieval bots made 1.1% of the fetches that did happen.³

Markup and feeds do matter in one place: the platforms. ChatGPT shows a Get Quote button on a business card "only when the business has an eligible service provider," and that provider supplies the feed.² For a business on no such platform, the feed is not yours to build. What reaches you is a question.

So the signal worth reading is not "AI can now find businesses." Discovery is narrow: SOCi's 2026 index found ChatGPT recommended 1.2% of the locations studied, per Near Media.¹¹ The signal is that when an assistant does pick you, it asks a short, predictable list of questions, and most small businesses do not answer them.

What do AI assistants ask a business?#

AI assistants that contact businesses ask about price and availability first. Google's help page lists what its automated calls do: "Book appointments," "Check restaurant wait times," and "Confirm price and availability of services and products."⁶ Gemini's Call for Me, rolling out from 24 September 2026 on Pixel 11 in the US, checks hours, availability and quotes, and identifies itself as an AI assistant at the start of each call.¹⁵ Yelp Assistant can "obtain estimates from professionals across 450+ service categories."⁷

The assistant then reports back to its user. Near Media tested Google's Ask for Me in July 2025: "In Mike's case, 3 out of 8 responded; in mine 2 out of 7." Results came "within an hour or less."⁵ A business that did not answer did not appear in the result. Nothing in the report suggests the user was told who stayed silent.

This is the same question people send. WeddingPro reports that 77% of couples say pricing is the most important factor they consider before contacting a vendor.⁹ The assistant is a faster, more literal version of the customer, and it does not wait.

Agent-ready, defined#

Agent-ready is a property of a business's answers, not its website code. We use this definition:

Each clause maps to a failure the data already shows. "Within minutes" answers the 26% of unanswered calls. "A price or range" answers the 48% that answered without one.¹ "A concrete next step" answers a broader gap: Pied Piper's 2026 home-services study, covering 3,211 locations, found only 28% offered an appointment when scheduling was the natural next step.¹⁴ "Facts the owner has written down" is what keeps an answer from becoming an invented price.

Speed carries a measured premium. Thumbtack reports pros who respond within 30 minutes are 25% more likely to get hired than those who respond within three hours (November 2025).¹⁰ An assistant compresses that window further, because it returns its shortlist in under an hour.

Items 1 to 6 are the six standard questions. Items 7 to 9 are what makes the six answerable in practice. For a fuller template of the facts behind items 1 to 6, see the facts sheet guide; for wording a price reply, see how to answer "what are your rates?"

What is an unanswered AI request worth?#

An unanswered AI request is worth the same as any other lost inquiry, because a customer is behind it. Invoca's July 2026 benchmarks found ChatGPT-referred calls had a 49% lead rate, at small volume, against 38% of answered calls being leads overall.¹² Invoca called it "a signal worth watching closely, not a channel worth diverting significant budget toward — yet."¹²

Some of these requests arrive disguised. Tygart Media reported in September 2026 that quote requests started in ChatGPT reach Yelp as billable leads "with no source tag."¹³ If you pay per lead, an unanswered AI request costs you twice: the fee and the job. Our guide to where ChatGPT and Yelp Assistant requests land covers each route.

How will agent requests evolve? Three scenarios#

Three scenarios bound how agent requests to small businesses could change over the next 18 months. Each has a leading indicator you can check without special tools.

The growth rate in the base case is already visible. Invoca recorded Google-AI pricing calls up 162% from September to October 2025 and 324% from October to November.¹ Growth from a small base says little about scale; it says the diffusion curve has started. If the base case holds, the constraint is not the assistant's capability. It is the supply of businesses that answer, which is a slower-moving input than any model release.

Consumer behaviour is moving faster than business readiness. BrightLocal found 45% of US consumers used AI tools for local recommendations in 2026, up from 6% in 2025, and 19% expect a same-day reply.⁸ The gap between those two curves is where requests go unanswered.

How could a business measure what assistants ask it?#

A business can measure its own agent requests with a simple log, and a vendor could run the same study across many businesses. The method matters more than any single figure, and no one has published a small-business dataset on it yet.

1. Tag the source. For 60 days, mark every inquiry that says it is from an AI assistant, or that came through a platform feature such as Get Quote or an automated call.
2. Code the question. Record which of the six standard questions it asked: price, availability, policy, area, minimums, next step.
3. Record the outcome. Answered or not, time to first reply, whether a price was given, whether a time was offered, and whether it booked.
4. Compare with people. Run the same four fields on inquiries from people over the same period.

This is a study Valet could run across the inquiries it answers, with owners' consent, counting assistant-originated requests and coding what they ask. We have not published such data, and the figures above come from the named third-party studies.

When to revisit this definition#

This definition should be revisited by 31 March 2027, or sooner if an assistant maker documents direct booking against a small business's own page, if Google publishes guidance specific to businesses receiving AI requests, or if a new study measures answer rates for written AI requests. The broader shipped-and-announced list is in agentic commerce in plain English, and the booking side is in Google AI Mode booking for off-platform businesses.

Where Valet fits#

Valet is one way to meet items 6 to 8 of the checklist for written requests. You forward an address such as info@ or bookings@ to Valet. It answers email, website forms, listing and marketplace lead emails, and requests on your free Valet page, including requests from AI assistants acting for a customer. It answers only from your own price list, calendar and policies, offers and holds real times, and sends anything off the list, including money questions and complaints, to you while telling the customer when you will reply. Valet does not answer phone calls. To see how your recent inquiries were handled, start with the free lead check: forward 10 to 20 of them and see which went unanswered and what they were worth.