How to use this report
Every check below is something an AI assistant does, in roughly this order, when a customer asks it for a local business and then asks it to act. The first stage is the one most SEO advice stops at: can the assistant find you and name you. The next three are where the money is in 2026, because the assistants now book, as we laid out in how AI agents book Denver service businesses. Each check has a pass condition, the evidence the assistant reads, and the fix. Score one point per pass. Be honest; the software will be. Keep a copy of your score with the date, because the useful number is not where you are today but how far you move in the next quarter, and which stage you are still stuck at when the next platform opens its booking rail to your trade.
Two sources anchor the thresholds. Whitespark’s 2026 Local Search Ranking Factors report, published November 6, 2025, surveyed 47 local search experts who scored 187 factors across the local pack, local organic, conversion and, for the first time, AI search visibility; its headline line is that in AI SEO “mentions (citations) are the new link,” and three of its top five AI factors are citation factors. BrightLocal’s 2026 Local Consumer Review Survey (1,002 U.S. adults, February 11, 2026) supplies the consumer thresholds that assistants are increasingly encoding: 31% require 4.5 stars, 47% skip businesses with fewer than 20 reviews, 74% want a review from the last three months, and 45% used an AI tool for a local recommendation, up from 6%.
Stage 1: Found (checks 1 to 4)
The assistant has to locate you and be sure the thing it found on Google is the same thing it found on your website and on Yelp. Four checks.
One name, one phone, one address or service area, everywhere
Pass: your business name, phone number, address (or hidden address with a stated service area), hours and primary category are identical on your website, Google Business Profile, Apple Business Connect, Bing Places, Yelp, Facebook and the directories you are listed in. What the agent reads: all of them, then compares. Ignite Visibility’s June 2026 analysis of Google’s agentic booking describes a consensus check across sources and a risk rule that drops a business with conflicting hours or services. Fix: pull every listing into one sheet, fix at the source, and claim Apple and Bing if you never did; we wrote the how-to in Bing Places and Apple Business Connect.
A dedicated page for each service you want to be booked for
Pass: “water heater replacement,” “furnace repair” and “sewer line” each have their own page with the service named in the title, heading, first paragraph and schema. What the agent reads: the page itself; Whitespark’s panel scored “dedicated page for each service” at 170, the second-highest AI visibility factor in the 2026 report. A single “Services” page that lists twelve things is one candidate for twelve questions, which is to say none. Fix: one page per service, each with price range, area, hours, proof and a booking link (checks 10 to 15 live on these pages).
AI crawlers are allowed in, and the facts are in the HTML
Pass: robots.txt does not block GPTBot, ClaudeBot, Google-Extended or PerplexityBot from the pages you want cited; your prices, hours and services are in the page text, not in an image or loaded by a script after the fact; the Search Console AI Overviews and AI Mode toggle is left on. What the agent reads: the raw HTML, usually without running JavaScript. Semrush’s April 2026 guidance on AI Mode booking puts “make sure AI crawlers can access your site” first on its list. Fix: check robots.txt and any firewall rule that blocks bots wholesale; view source on your top service page and confirm the price range is visible there. If you are tempted by the Search Console opt-out, read why we tell most local businesses to leave it on.
You appear on at least three third-party lists or industry domains
Pass: a search for “best [your trade] Denver” in an assistant returns at least three lists, directories or industry sites that include you. What the agent reads: curated lists and prominent domains; Whitespark’s top AI factor at 179 is “presence of business on expert curated best-of and similar lists,” and “prominence on key industry-relevant domains” is third at 167. Fix: earn the lists you can (chamber, trade association, neighborhood guides, local press), and make sure your directory listings are complete and verified rather than skeletal. Our post on which review sites AI assistants read ranks the sources by trade.
Stage 2: Trusted (checks 5 to 9)
Found is not recommended. The assistant is about to put its credibility behind a business the customer will never research themselves, so it checks for the things a cautious friend would check.
Rating at or above 4.5, with at least 20 reviews, aiming at 50
Pass: 4.5 or better on Google with 20 or more reviews; 50 is where the AI summaries have enough material to be specific. What the agent reads: count, rating, and increasingly the text. BrightLocal 2026: 31% of consumers require 4.5 stars (up from 17%), 47% will not use a business with fewer than 20 reviews, and 82% read Google’s AI-generated review summaries. Fix: ask every customer at the close of the job, one time, with a direct link; no incentives (Google prohibits them, and its July 2026 review-snippet guidance says structured data must not represent incentivized reviews). The full plan is in the 2026 review numbers for Denver.
A review from the last 90 days, and a reply on every review
Pass: newest review under 90 days old; every review, good and bad, has a non-templated reply. What the agent reads: dates and reply text; recency is a proxy for “still open and still good,” and replies are evidence a human is there. BrightLocal: 74% want reviews from the last three months, 89% expect a reply, 50% are put off by templated ones. Fix: a review ask built into the job close rather than a quarterly campaign; a 90-second reply that names the specific job.
License, insurance and years in business are visible on the site
Pass: license number (where your trade is licensed in Colorado or Denver), proof of insurance, and a founding year appear in the footer or on the About page in text, not only in a badge image. What the agent reads: text it can quote back to the customer: “licensed, insured, in business since 2009.” Fix: add a one-line credentials strip to the footer and the same facts to Organization schema (foundingDate, and the license in a description or identifier field).
The site is secure, fast enough and actually loads
Pass: HTTPS with a valid certificate, no mixed-content warnings, the home and top service pages load in under about three seconds on a phone, and nothing on the page is broken. What the agent reads: whether it can fetch the page at all, and a handful of trust signals that overlap with what people check before they hire. Our sister directory Vettley publishes free business signal reports that check exactly these public signals, a secure site that loads, a real address and phone, time online, so people can check a business before hiring it; run your own and see what an agent sees. Fix: renew the certificate, fix the broken pages, and treat speed as a trust signal, not a vanity score (see Core Web Vitals in 2026).
No contradictions an agent can catch
Pass: the site does not claim 24/7 while Google says 8 to 5; the price range on the service page matches the one in the FAQ; the service area on the About page matches the one in your profile; the phone number on the ad landing page is the same number. What the agent reads: everything, cross-referenced. The risk rule in Google’s agentic flow is explicitly triggered by conflicting hours or services. Fix: a single source of truth (one document with every fact) and a quarterly pass to re-sync the site, profile and listings against it.
Stage 3: Quoted (checks 10 to 12)
This is the stage most local businesses fail, and it is the one that decides whether the customer hears your name with a number next to it or hears a competitor’s.
A price or price range per service that an assistant can repeat
Pass: each service page carries a sentence like “furnace repair visits in Denver typically run $150 to $450; a full replacement runs $4,500 to $9,000 depending on efficiency and ductwork.” What the agent reads: that sentence, plus Offer or priceRange schema if present. “Call for a quote” is not an answer; it is a dead end, and the agent moves to a business that can answer. Fix: ranges with the variables named. A range is not a promise of a number; it is the thing the assistant reads out loud. If you are worried about being held to it, say what moves it.
Hours and service area the agent can read as data
Pass: openingHoursSpecification and areaServed (the suburbs and ZIPs you drive to, named) in LocalBusiness schema, matching the visible text and the profile. What the agent reads: schema first, visible text second. A customer in Highlands Ranch asking at 9 pm needs the agent to know you work there and whether you take evening calls. Fix: list the service area as towns and ZIPs rather than “the Denver metro,” and keep holiday hours current on the profile.
The job the customer describes maps to a service you name
Pass: the words customers use (“water heater died,” “no heat,” “roof leaking after hail”) appear on the relevant service page, next to the service name you use. What the agent reads: the match between the customer’s sentence and your page. A page that only says “hydronic system diagnostics” will not be matched to “my radiators are cold.” Fix: a short “symptoms we fix” list on each service page in plain words, and a question-shaped FAQ with FAQ schema.
Stage 4: Booked (checks 13 to 15)
The last three checks are new to most businesses and they are the ones that turn a recommendation into a job on the calendar.
Live availability an assistant can read
Pass: a scheduler that exposes open slots (a Reserve with Google partner such as Booksy, OpenTable, SevenRooms, Mindbody, Boulevard, or a trades scheduler such as Jobber, Housecall Pro or ServiceTitan with online booking on), or an availability endpoint. What the agent reads: the partner feed or the endpoint. As of October 2026 ChatGPT reads OpenTable, Resy, Yelp and Booksy; Google AI Mode reads its Reserve with Google partners. Fix: turn online booking on with real slots; ask your vendor directly whether and how they expose availability to Google and to the ChatGPT app directory.
A booking method that is not a static form
Pass: the primary call to action on each service page is a booking link with times, a Reserve with Google button, or a tool an assistant can call with the customer’s approval; ReserveAction is declared in potentialAction in the page schema. What the agent reads: whether it can see what happens after the click. A contact form is a wall; the agent cannot confirm anything on the other side. Fix: keep the form for people who want it, but make the booking link primary, and add the schema. For trades, a private booking endpoint (an MCP server, the open standard the assistants use to connect to tools) answers area, price and slots and takes the booking; we explain what that is in plain English in agent cards, MCP and llms.txt for local business.
Someone confirms within a minute, day or night
Pass: every booking or lead gets a confirmation in seconds and a human (or a disclosed AI assistant on your behalf) within a minute, including at 9 pm on a Sunday. What the agent reads: nothing directly; the customer reads it, and the review they leave teaches the next agent. Fix: route bookings to a phone that is answered, with an instant reply for the hours it is not. This is the same discipline we built lead response time around, and it applies double to a customer who was comfortable letting software book them.
Scoring and what each band means
| Score | What it means | What an assistant does with you today | First fix |
|---|---|---|---|
| 0 to 5 | Not reliably found | May name you from the map pack; cannot verify or quote you; will not offer to book | Checks 1 and 2: one identity, one page per service |
| 6 to 8 | Found, not trusted | Names you with caveats; prefers a competitor with more recent reviews and visible credentials | Checks 5 to 7: reviews and credentials |
| 9 to 11 | Trusted, not quotable | Recommends you, then tells the customer to call for pricing; many customers stop there | Check 10: price ranges with variables |
| 12 to 13 | Quotable, not bookable | Reads your range aloud, then sends the customer to a form or a phone number | Checks 13 and 14: availability and a booking method |
| 14 to 15 | Bookable | Can find, verify, quote and book you without a phone call; this is the short list | Check 15: keep the response discipline |
When we run this on Denver service businesses the median is around 6. Established companies with strong reviews often sit at 8 or 9 and are surprised; they have been winning on reputation while failing on readability. The jump from 9 to 12 is mostly writing (price ranges, service pages, a credentials strip) and the jump from 12 to 15 is mostly configuration (a scheduler, a booking link, the schema). Neither requires a redesign.
What the assistant actually says at each band
The score is abstract until you hear it. These are the shapes of answer we see from the assistants, paraphrased, when a customer in a Denver suburb asks for a service and then asks to book it.
At 0 to 5: “Here are a few highly rated options in Arvada,” followed by three names, none of them yours, pulled from the map pack and a review site. You are not in the conversation and the customer has no idea you exist.
At 6 to 8: “Your Company is also listed nearby, though I couldn’t confirm its current hours and the most recent review I can see is from last spring.” You are named, with a hedge, and the hedge is doing the work of a competitor’s ad.
At 9 to 11: “Your Company has strong recent reviews and is licensed; they don’t list pricing, so you’ll want to call them for a quote. Competitor B lists furnace repair at $150 to $450.” This is the most painful band, because you earned the recommendation and lost the job on one missing sentence.
At 12 to 13: “Your Company quotes $150 to $450 for a repair visit and serves Arvada. I can’t see their availability, but here’s their contact page.” Half the customers will follow that link. The other half will say “just book whoever has an opening,” and that will be someone else.
At 14 to 15: “Your Company can come Tuesday at 8 am or Wednesday at 10 am for a diagnostic at $150 to $450. Want me to book Tuesday?” That is the sentence this whole checklist exists to produce.
The failures we see most often in Denver
Running this on local companies for a year, the same handful of failures recur, and they are rarely the ones owners expect. Hours are the most common contradiction: a profile set to 24/7 by a marketing company in 2023 and a website that says 7 to 6. Service areas are the second: “Denver metro” on the site, a 25-mile radius on Google, and a crew that actually will not go past C-470 on a Friday. Price pages are the third: a “pricing” page that exists in the navigation and says “every job is different, call us.” Fourth is the review gap after a good year: a hundred reviews, a 4.8, and nothing since June because the ask was tied to one office manager who left. Fifth, and newest, is the form: a beautiful new site whose every button leads to the same four-field form, which the agent reads as a locked door. Every one of these is a one-afternoon fix once it has been named, and naming it is what the score is for.
The schema that does the most work
You do not need every schema type. You need the handful that carry the five inputs an agent is reading for. Here is the minimum set we put on a service page, in order of effect.
| Type / property | Carries | Check it serves |
|---|---|---|
| LocalBusiness (or the specific subtype: Plumber, HVACBusiness, RoofingContractor, Dentist) | name, telephone, address or areaServed, openingHoursSpecification, priceRange, foundingDate, sameAs to every profile | 1, 7, 11 |
| Service with provider and areaServed | the named service, where you do it, offers with a price range | 2, 10, 11 |
| Offer / PriceSpecification (minPrice, maxPrice, priceCurrency) | the range the agent repeats | 10 |
| potentialAction: ReserveAction with a target urlTemplate | the booking link the agent follows | 14 |
| FAQPage | customer-language questions mapped to your services and prices | 12 |
| AggregateRating (only from real, visible reviews on your site; never copied from Google) | rating and count where Google’s guidelines allow it | 5 |
Two cautions. Google’s review-snippet guidelines, updated July 2026, are explicit that structured data must represent real reviews on your own site and must not represent incentivized or undisclosed ones; marking up your Google rating on your website is a violation that can cost you rich results and, more to the point, an agent’s trust when it notices the mismatch. And schema that disagrees with the visible page is worse than none; check 9 applies to your own markup.
Beyond schema, three small text files are becoming the place agents look for a business’s own description of itself: llms.txt (a plain-language summary for language models), an agent card (the A2A protocol reached version 1.0 on March 12, 2026 with cryptographically signed Agent Cards so an agent can confirm a card was issued by the domain owner), and, in Chrome’s early preview, WebMCP, which lets a page register the actions it supports, such as “book an appointment,” so a browser agent does not have to guess at your buttons. None of these is required to pass the fifteen checks. All of them make passing easier, and we cover which ones are worth a small business’s time in the companion post.
Order of operations for a Denver business
If you fail more than five checks, do not try to fix them in numerical order. Fix them in the order that unlocks the next stage.
- Day one: the single source of truth. One document with your exact name, phone, address or service area (towns and ZIPs), hours, category, services, price ranges with variables, license, insurance and founding year. Everything else is copied from this.
- Week one: consensus. Re-sync Google, Apple, Bing, Yelp, Facebook, the directories and your footer to the document. Confirm robots.txt is not blocking the AI crawlers. Confirm the Search Console toggle is on.
- Week two: service pages. One page per bookable service with the range, the symptoms list, the area, the credentials strip and the schema set above. This is foundational onsite work; if you want it done task by task, it is what the SEO à la carte menu sells at a fixed price per task.
- Week three: reviews and replies. Start the job-close ask, reply to everything, and keep a review newer than 90 days from here on.
- Week four: booking. Online booking on in the scheduler, booking link primary on every service page, ReserveAction in the schema, instant confirmation wired, and the “Bookings and inquiries from customers” toggle on in your Google Business Profile advanced settings.
- Then: test it. Ask ChatGPT, Gemini, Claude and Google AI Mode, in plain words, to find a business like yours in your suburb, tell you what it costs and book it. Save the transcripts. Re-run quarterly. That transcript is the only readiness report that matters, and it is free.
If you would rather have it scored for you, our AI readiness audit page explains the free ten-second scan, the $297 written audit and the $750 done-for-you option, and which one fits a given score. The free audit includes the readiness pass as well. And the AI assistant in the header of this page will tell you, right now, what it can and cannot find out about your business from public sources, which is a decent proxy for what the big assistants see.
Want your 15 checks scored by a person?
The free audit runs this checklist on your business and tells you which stage you are stuck at and the first three fixes. You keep the findings either way.
Get the free audit → Call 1-800-481-8638Questions readers ask
What does AI readiness mean for a local business?
Whether an AI assistant can complete four stages with your business: find it with one consistent identity, trust it from reviews, credentials and a working site, quote it from a published price range, and book it from live availability through a method that is not a static form. This checklist scores those four stages across 15 checks.
How many Google reviews does an AI assistant want to see?
There is no published threshold inside the assistants, but they increasingly encode consumer behavior, and BrightLocal’s 2026 survey found 47% of people skip businesses with fewer than 20 reviews and 31% require 4.5 stars. Twenty is the first target, 50 is where Google’s AI review summaries become specific about you, and a review newer than 90 days matters as much as the count.
Do I need schema markup to be bookable by AI?
Not strictly, but it is the cheapest way to make the five inputs unambiguous. The minimum set is LocalBusiness (or its subtype) with hours, area and priceRange, a Service with an Offer range, a ReserveAction pointing at your booking link, and FAQPage. Schema must match the visible page; a mismatch is a trust failure.
Should I block AI crawlers to protect my content?
For a local service business, almost never. Blocking GPTBot, ClaudeBot or Google-Extended removes you from the answers your customers are asking for, and the Search Console AI toggle removes you from AI Overviews and AI Mode. The content on a service page exists to be quoted; let it be.
What is the fastest way to raise a low score?
Fix identity first (check 1), then build one page per service with a price range (checks 2 and 10). Those three moves take a business from “not reliably found” to “quotable” and are mostly writing. Availability and a booking method (13 and 14) are configuration and come next.
Is a free AI readiness scan accurate?
A scan reads your public pages and grades the readable signals in seconds; it is accurate about what is missing from your site and listings, and it cannot see what happens inside your scheduler or how fast you answer. Use it to find the gaps, then confirm with the transcript test: ask an assistant to book you and read what it says.
What should a Denver business do about hidden addresses?
Hiding a home address on Google is correct for service-area businesses. Compensate by stating the service area as towns and ZIPs on the site and in schema, showing license and insurance in text, and keeping reviews recent, so the agent has enough to verify without an address.
Sources
- Whitespark, 2026 Local Search Ranking Factors (published November 6, 2025; 47 experts, 187 factors)
- BrightLocal, Local Consumer Review Survey 2026 (February 11, 2026; 1,002 U.S. adults)
- Google Search Central, review snippet structured data guidelines (updated July 2026)
- Google Business Profile Help, automated calls and texts from Google (Bookings and inquiries toggle)
- Ignite Visibility, Google agentic booking analysis (June 4, 2026): consensus check and risk rule
- Semrush, worldwide agentic restaurant booking via AI Mode (April 15, 2026): crawler access guidance
- OpenAI Help Center, ChatGPT release notes (August 10, 2026: OpenTable, Resy, Yelp)
- Booksy, getting booked through ChatGPT (October 20, 2026)
- Linux Foundation, A2A protocol first-year milestones and v1.0 signed Agent Cards (April 9, 2026)
- VentureBeat, Chrome ships WebMCP in early preview
- Google Maps, policy-violating reviews removed in 2025 (292M)