If an assistant describes a service you do not offer, or sends a buyer to a broken contact form, an AI readiness review should help you locate and fix the problem. A useful review connects what customers can learn about your business with the next step they take when they want to contact you.
A technical scan can contribute evidence, while the business still needs to decide which services to promote and approve the conditions its staff explain every day. Bringing those decisions into the review makes the findings easier to prioritize and the resulting changes easier to check.
For a growing business, we would start with one service and one customer journey. You can expand once you have a method that produces clear findings and completed work, and by then each new improvement arrives with an owner and a place in the order.
Begin with the questions a buyer needs answered
Pick a service that matters to the business and list what a potential customer needs to know before inquiring. Use the questions your team receives, where you have them, and label any additional questions as your own hypotheses.
For example, a property inspection business might need to explain which property types it covers and what the visit includes, along with when the customer receives the report. The answer may also depend on location or the kind of inspection required.
Keep the first question set small enough to review carefully. Include a question about the business by name and a broader question from someone who has never heard of it, since these examine two situations: whether a tool describes you accurately and whether it includes you in a relevant discovery answer.
Write down the answers the business considers correct, with the page or internal source that supports each fact. A reviewer needs that reference before labeling an external answer inaccurate.
This exercise can reveal missing decisions straight away. If two people on your team give different explanations of the same service, settle on the offer before asking a writer or an assistant to describe it.
Review the public facts your business controls
Inspect the relevant website pages and important business profiles. Look for contradictions that could affect a decision, such as a service listed in one place but retired elsewhere.
For an eligible local business, Google provides ways to describe services on a Business Profile. Those details need to agree with the current offer on the website, and updating one does not establish that the other was corrected.
Create a small facts sheet with the approved wording and the person responsible, along with the places it appears. This gives future changes a clear path, so when opening hours change, the owner can work from the sheet and skip trying to remember every place someone once copied them.
Do not use the review to add claims the business cannot support. If you do not publish fixed prices because the job requires an assessment, explain how estimates work. Clear conditions help a buyer more than a made-up number inserted to satisfy an imagined search requirement.
The Markdown for Agents guide includes a comparison exercise for checking that conditions remain understandable in a readable text copy.
Check whether the pages can be found and read
The reviewer should inspect the important page itself, since the homepage says little about the state of a service page. Keep the address and the result with the finding.
Google’s URL Inspection documentation distinguishes information about an indexed page from a test of the current live page. That distinction matters after a repair, because the live page may be corrected while Google’s stored information still reflects an earlier version.
A report should say which observation it used. “The current page returned an error” and “the indexed version is older” lead to different actions.
Read the accessible text too. A successful response can still leave out the service explanation a buyer needs, while a single failed fetch may need investigation before the reviewer calls the site inaccessible.
If the review includes a tool score, ask how each relevant result maps to the business journey. The Cloudflare readiness score guide explains how to separate repairs from optional capabilities.
Use assistant answers as dated observations
Ask the saved customer questions in the environments you intend to monitor and keep the responses. Record the platform, date and any relevant location setting, along with visible sources where available.
Read beyond whether the business name appears. An answer may mention you while getting a service limitation wrong, and another may leave you out of a broad recommendation yet describe the business accurately when asked directly.
Those are separate findings. One suggests correcting a fact or its source, while the other may need investigation into the comparison itself and the evidence available about your business.
Repeatable questions help you observe changes over time. They do not prove what every customer sees, and changing the question at the same time as the website makes the result harder to interpret.
Keep an unexpected answer even when it is inconvenient. A review that records only the flattering examples gives you little help deciding what to fix.
Follow through to the inquiry
Choose a labeled test inquiry and complete the normal customer route. Check the confirmation the visitor sees, then find the received request in the business system.
Look at who receives it and whether that person has enough information to respond. If an inquiry arrives without the selected service or with no assigned owner, the website can attract the right person and still create avoidable work.
This part of the review should use test details the team has approved and avoid creating a live appointment or payment. Clean up the test records afterward so they do not distort reporting.
For a connection that needs work, the native integration, Zapier and custom comparison gives you a practical way to evaluate the options. A readiness review can identify the failed step, and the repair for each step can then go to whichever kind of project suits it.
Turn findings into a first round of work
Consider this hypothetical review. The main service page omits a commercial inspection option and the inquiry form drops the property-type field, while the site also lacks an optional assistant connection.
We would repair the missing inquiry field and publish the approved service information first, since both address a concrete customer problem. The assistant connection needs its own reason before it joins the build list.
| Finding | Evidence | First action | Completion check |
|---|---|---|---|
| Commercial service is missing | Current service page and approved offer | Add the commercial scope and exclusions | Owner approves the published explanation |
| Property type is lost | Test submission and received CRM record | Repair the field mapping | A new test arrives with the selected value |
| External profile has old hours | Profile view compared with approved hours | Correct the profile | Public details show the intended hours |
| Assistant answer is incomplete | Saved question, response and visible source | Investigate whether a relevant source lacks detail | Repair verified gaps and repeat the observation later |
| Optional connection is absent | Technical scan | Identify a customer task it would serve | A documented build or defer decision |
Add an owner and a target date to each accepted action, and record dependencies as well. A profile correction may need account access, while a service rewrite needs the business to settle its wording.
Keep the size of the first round realistic. Completing two useful repairs gives you stronger evidence of progress than approving a large list while the people needed for the work remain unavailable.
Judge delivery and outcomes on different schedules
You can verify a repaired field as soon as it is deployed, and approved wording as soon as the page is published. Search visibility and assistant answers need observation over time, and several influences may change together.
Google’s official guidance for generative search emphasizes useful content and ordinary SEO practices. It also makes clear that meeting the requirements does not guarantee inclusion, so a review should avoid promising a recommendation in exchange for a technical repair.
At the end, you should have a facts sheet, saved evidence and a prioritized work list. After implementation, each completed item should have a check someone can repeat, while unresolved items remain visible with their next step.
That gives you a useful basis for deciding what to do next month. You can see which customer problems were resolved and which observations still need investigation, then judge whether new work would serve the business you are trying to grow.