Example report: Example Roofing Company (example)

Example report. The business is fictional; the findings are typical of what our audits find.

What a CiteMy.ai audit finds, and what we build

An example audit of exampleroofing.example, a fictional roofer in the East Valley: what our AI business engine found, and the work that follows, new data first and cleanup after. This is where every client's work starts.

Who AI recommends

We asked ChatGPT and Perplexity: “Who is the best roofer in Mesa, AZ?” and 2 more questions a customer would ask.

They named: Roofer A (example), Roofer B (example), Roofer C (example).

You were named in 0 of 6 answers.

Example. A real report shows the businesses the AI assistants named in its own run.

AI visibility score, for this example

34/ 100

22 findings in 7 areas and an implementation plan of 33 steps. Example figures for a fictional business.

Five areas, each out of 100

  • Data engines can quotellms.txt, structured facts, datasets12
  • Crawler accessrobots.txt, sitemap, rendering38
  • Business identityOne entity in the schema and across listings30
  • ContentNamed authors, dates, one question per page45
  • Local presenceListings, hours, service areas52

In this example the lowest area is data the engines can quote, which is where the work starts.

What we found

By area, worst first, in plain words.

Data the engines can quote

  • High

    No data to cite anywhere on the site

    There is no price, spec, warranty term, service-area list, or timeline the engine can pull out and attribute. The pages read like brochures; the engine has nothing to name a source for.

  • High

    Content not structured for AI ingestion

    The pages are plain HTML paragraphs with no schema.org types on the facts they mention. An engine that finds the words has no way to tell what is a price, what is a service area, and what is a headline.

  • High

    No llms.txt

    The site has no /llms.txt, so an AI engine gets no plain-text map of the services, prices, datasets, and answers the site offers.

  • High

    Prices, service areas, and warranty terms only in prose

    Repair prices appear in a paragraph on one page. None of it is marked up as Offer, areaServed, or additionalProperty, and there is no dataset an engine can cite.

  • Medium

    No datasets in Google Search Console or dataset search

    The site has no Dataset schema, no CSV or JSON downloads, and nothing indexed by Google's dataset search or the open data portals engines pull from. First-party data is invisible to the pipelines that carry it.

  • Medium

    Customer questions answered without numbers

    The FAQ answers “How much does a tile roof repair cost?” with “It depends.” The business has the numbers from its own jobs; the page does not use them.

Business identity and schema

  • High

    Business nodes that disagree

    A theme, an SEO plugin, and a reviews widget each write their own LocalBusiness block, with no shared @id and two different phone numbers. An engine cannot decide which one is the business.

  • Medium

    Hours, areaServed, and geo missing

    None of the business nodes says when the business is open, which cities it serves, or where it is.

AI crawler access

  • High

    PerplexityBot blocked, and a Crawl-delay for every crawler

    robots.txt shuts out PerplexityBot and sets Crawl-delay: 10 for everyone, which slows every crawler that honors it.

  • Medium

    Sitemap stale

    The XML sitemap was last generated two years before the audit and lists pages that no longer exist.

  • Medium

    Key content rendered only by JavaScript

    Service pages load their prices and service areas with JavaScript, so a crawler that does not run scripts sees an empty section.

  • Low

    No Bing Webmaster Tools or IndexNow

    Nothing tells the engines when a page changes; they find out whenever they next crawl.

  • Low

    No way for an AI agent to request a quote

    No AI Catalog, MCP Server Card, or /openapi.json, so an assistant acting for a homeowner has no direct way to ask this business for a quote.

Proof of real jobs

  • Medium

    City pages with no job in them

    Six city pages share one template with the city name swapped: no price, no photo of work done there, no finished project named. That is the thin, templated pattern engines discount.

  • Medium

    No project pages

    There is a photo gallery, but no page for any single job: nothing an assistant can cite that says what was installed, where, and for roughly how much.

  • Medium

    No rebate or energy-savings page

    Cool-roof rebates and what a reflective roof saves on a summer bill are mentioned once, in passing, on the foam page. There is no page that lists each program with its amount and its end date.

  • Low

    Brands named, but no certification and no brand page

    The site says it installs GAF and Owens Corning shingles but states no installer certification with either, and has no page about either brand.

Content

  • Medium

    No author or dates on posts

    Blog posts carry no named author and no published or updated date.

  • Medium

    Nothing published in the last 90 days

    The newest dated post is eight months old.

Google profile and measurement

  • Medium

    Google Business Profile never linked from the site

    No link and no sameAs points to the Google profile, so nothing ties the site to the profile its reviews and hours come from.

  • Low

    No analytics tag

    No GA4 or Tag Manager tag on any page, so visits and leads from ChatGPT, Perplexity, Claude, and Gemini are not counted.

Directory listings

  • Medium

    Old address and phone on two directories

    Two directory listings still show the address and phone from before the business moved.

The implementation plan

Every free audit ends with one: everything the audit found plus the work every client gets, in five groups from quick wins to ongoing work, with the hours for each step and how much each group moves the needle. This is the example business's plan, made by the same rules a real audit uses.

33

steps, 29 of them found on your site

169–337

hours of one-time work

31–57

hours every month after that

1. Quick wins (3)

You can do this · 4.5 to 9 hours

Fixes you can make yourself, mostly from logins you already have.

ImpactMoves the needle a lotThese are the basic facts every AI engine checks first: one name, address, and phone everywhere, and the profiles they read. Small effort, big effect.
  1. Claim your ChatGPT business listing (Business Hub) and keep its website, description, category, and photos accurateOpenAI opened Business Hub on November 10, 2026: a business can now claim the listing ChatGPT shows for it and see how often ChatGPT mentions it. Claiming does not decide placement.30 minutes to 1 hourFound on your site
  2. Update every third-party listing to the primary name, address, and phone; any listing for a second line of business must be clearly distinctBefore merging or retiring anything, ask what else the business does: a second phone, address, license, or listing can be deliberate segmentation by line of business, and then it must be clearly distinct, not removed. Found: Old address and phone on two directories.2 to 4 hoursFound on your site
  3. Fill the Google Business Profile: primary category, services, products with prices, photos, Q&A, and a UTM-tagged website linkA complete profile is the local entity record every engine reads.2 to 4 hoursFound on your site

2. Technical but doable (8)

You or your web person · 10 to 20 hours

Changes in your site's settings and files. Doable with patience, or a short job for your web person.

ImpactHelps, but a slower return on your timeEach fix helps the engines read your site, but most of the hours go into learning the tools before the fix itself.
  1. Register Bing Webmaster Tools and enable IndexNowChatGPT search runs on Bing's index.1 to 2 hoursFound on your site
  2. Remove the Crawl-delay line from robots.txt, or set it to 1A long crawl delay caps how many pages a crawler refreshes per day. Found: A Crawl-delay for every crawler.1 to 2 hoursFound on your site
  3. Replace the conflicting business nodes with one business node with a stable @id, referenced from every pageEngines resolve a business by consistency; several nodes with different phones and addresses lower confidence. Found: Business nodes that disagree.3 to 6 hoursFound on your site
  4. Let GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, and Applebot fetch the site, and lift any Google-Extended opt-outAn engine that cannot fetch a page cannot cite it. Found: PerplexityBot blocked.1 to 2 hoursFound on your site
  5. We set up GA4 on every page with AI assistants as their own channel and every form submit counted as a conversionWithout a tag, visits and leads from ChatGPT, Perplexity, Claude, and Gemini cannot be counted. Found: No analytics tag.1 to 2 hoursFound on your site
  6. Link the Google Business Profile from the site's footer and contact page, and add it to the business node's sameAs and hasMapThe link is how an engine ties the site to the profile its reviews and hours come from. Found: Google Business Profile never linked from the site.1 to 2 hoursFound on your site
  7. Publish an llms.txt listing the key pages, facts, and datasetsllms.txt tells an assistant what to read first. Found: No llms.txt.1 to 2 hoursFound on your site
  8. Add chatgpt.com, perplexity.ai, claude.ai, gemini.google.com, and copilot.microsoft.com as an AI channel group in GA4Shows visits and leads from assistants.1 to 2 hoursFound on your site

3. Big jobs (5)

Needs your web person · 52 to 104 hours

New pages, page templates, and content rewrites. Real hours from someone who builds websites.

ImpactSolid, but slowThis is the content the engines quote, but it is a lot of hours before any of it shows up in an answer.
  1. Publish a spec table on every product page, with a key-facts block of three to five sentences with numbers at the topSpec questions are answered from sites that print the number in a table. Found: No data to cite anywhere on the site.20 to 40 hoursFound on your site
  2. Create author pages for two or three real people and put a byline and visible date on every postNamed, dated authorship is a trust signal and our content rule. Found: No author or dates on posts.8 to 16 hoursFound on your site
  3. Rebuild each city page around the real jobs done there: a price range, before-and-after photos, and the finished projects in that city, linkedCity pages without a job in them are thin, templated text, the doorway pattern engines discount. Found: City pages with no job in them.8 to 16 hoursFound on your site
  4. Write honest head-to-head pages against the main competitors and a fair 'companies in the metro' list page"X vs Y" is one of the most common assistant query shapes.8 to 16 hoursFound on your site
  5. Add one FAQPage to each product page with questions from real sales calls, answered with numbersQuestions customers actually ask, answered on the product page, match how people phrase assistant queries.8 to 16 hoursFound on your site

4. Specialist work (8)

Hard to get right without experience · 102 to 204 hours

The data, schema, and pages AI engines quote. All of it can be done without us; most do-it-yourself attempts stall here, because small mistakes make the engines ignore the work.

ImpactMoves the needle a lot, and builds a moatFew competitors do this work. Once it is live and the engines trust it, a business starting later has to catch up on all of it.
  1. We build: one page per product line and brand you install, with specs, warranty, energy ratings, and Product schemaSpec, warranty, and brand questions are answered from pages that print the numbers for one product. Found: Brands named, but no brand page.30 to 60 hoursFound on your site
  2. We build: one page per completed job, with the photos, the product installed, the neighborhood, and the price rangeA finished job with a place, a product, and a price is proof an assistant can point to when it recommends a contractor. Found: No project pages.10 to 20 hoursFound on your site
  3. Build a /data/ section with a dataset page template: HTML table, CSV and JSON downloads, methodology, update date, CC BY 4.0, Dataset schemaFirst-party datasets are the part competitors do not copy. Found: No datasets in Google Search Console or dataset search.30 to 60 hoursFound on your site
  4. We build: rebate and energy-savings pages with the current utility rebates and federal credits (amounts and end dates) and what the work saves on a bill, kept currentRebate and savings questions are answered from pages that list each program with its amount and its end date. Found: No rebate or energy-savings page.10 to 20 hoursFound on your site
  5. We build: a manufacturer page naming the brands you install and your dealer or installer certifications, and get you listed in each manufacturer's dealer locator under the site's exact name"Who installs <brand> near me" is answered from dealer locators and from installer pages that name the brand. Found: Brands named, but no certification.10 to 20 hoursFound on your site
  6. Give AI agents a direct way to reach you for the person they work for: an MCP server and web API (Agent Doorbell), signposted from /.well-known/ai-catalog.json, an MCP Server Card, /openapi.json, and llms.txtAssistants that act for a customer look for these files to find a business they can send a request to. Found: No way for an AI agent to request a quote.4 to 8 hoursFound on your site
  7. Create a Wikidata item for the company with founding year, location, website, and license numberWikidata is a source assistants already trust for entity facts.4 to 8 hoursFound on your site
  8. Quick-facts page: one page with your key facts, where we correct what AI engines get wrong about youEvery client gets one: it is where the facts AI assistants get wrong are corrected.4 to 8 hoursEvery client gets this

5. Ongoing work (9)

Every week and every month · 31 to 57 hours a month

Work that never finishes. AI answers change every week, and so does what they read.

ImpactKeeps you named, and widens the gapThe businesses that keep feeding the engines new, consistent facts are the ones that stay in the answers. Every month of it adds to a record a competitor would have to match.
  1. Run the prompt-panel baseline before any change ships, then monthlyWithout a baseline there is no before and after.3 to 5 hours a monthFound on your site
  2. We publish at least 4 pieces a month, each signed by a named person on your staff and answering one customer question with a number from your own jobsEach answered question is one more page an engine can cite, and a steady date line shows the business is active. Found: Nothing published in the last 90 days.12 to 20 hours a monthFound on your site
  3. We track your Google Maps rankings for your main services in each city you serve, every weekA crawl of the site cannot measure local rankings, so tracking starts on day one and the weekly report shows it.1 to 2 hours a monthFound on your site
  4. Reply to every Google review, each approved by you before it postsReview volume and replies are the local signals assistants quote.1 to 3 hours a monthFound on your site
  5. Answer homeowner threads in the metro and home-improvement subreddits as a named employee, with the affiliation disclosedPerplexity and ChatGPT lean heavily on Reddit for local recommendations.2 to 4 hours a monthFound on your site
  6. Pitch one story per quarter to local news, built on a published datasetA news story that cites the dataset is an independent source for the same facts.1 to 3 hours a monthFound on your site
  7. AI Citation Defense: every week we ask the AI engines your customers' questions and track who they name instead of youOngoing: every week, for as long as you are a client. A crawl of your site cannot see what the engines say about you; the weekly measurement can.6 to 10 hours a monthEvery client gets this
  8. Job videosVideos start once your job software (Housecall Pro or Jobber) is connected; a crawl of your site cannot see it.3 to 6 hours a monthEvery client gets this
  9. Distribution where AI engines read: sending every new page and dataset to the places AI engines read, as soon as it publishesOngoing: every new page and dataset is sent out as it publishes.2 to 4 hours a monthEvery client gets this

Hours are typical for this kind of work, not a quote. Ratings describe what this kind of work usually does, not a promise for your business.

Don't have time for this? We can do it for you.

No contract. Cancel any time. Within 30 days of receiving the logins we need, we ship your quick wins and start your weekly AI tracking, or that month is free.

What CiteMy.ai would build for this business

Every CiteMy.ai deliverable this example audit found missing or weak, and the work every client gets. All of it is part of the monthly service, month to month, no contract.

  • Not on your site

    Rebate and energy-savings pages

    We build rebate and energy-savings pages: the current utility rebates and federal credits with amounts and end dates, and what the work saves on a bill.

  • Not on your site

    llms.txt

    We write and maintain your llms.txt and llms-full.txt: a plain-text map of your services, prices, datasets, and answers for AI assistants.

  • Not on your site

    Datasets

    We publish your own numbers as datasets (prices, specs, local rules) with CSV and JSON downloads and Dataset schema.

  • Not on your site

    Analytics setup

    We set up GA4 with AI assistants (ChatGPT, Perplexity, Claude, Gemini, Copilot) as their own channel and every form submit counted as a conversion.

  • Not on your site

    Agent Doorbell

    We set up Agent Doorbell: a direct way for AI agents to request a quote from you for their user, with an MCP server, a web API, and the discovery files agents look for (an AI Catalog, an MCP Server Card, and /openapi.json). Requests land in your inbox or job software, marked as an AI agent.

  • On your site, needs work

    Service-area pages from real jobs

    We build one service-area page per city you serve, from your real jobs there: the work done, price ranges, before-and-after photos, and job details.

  • On your site, needs work

    Project pages

    We build one page per completed job: the photos, the product installed, the neighborhood, and the price range.

  • On your site, needs work

    Product and brand pages

    We build a page for each product line and brand you install: specs, warranty, energy ratings, and Product schema.

  • On your site, needs work

    Manufacturer page and dealer-locator listings

    We build a manufacturer page naming the brands you install and your dealer or installer certifications, check your listing in each manufacturer's dealer locator, and give you the exact profile copy to submit in each portal (you submit it; we never ask for your password), then verify it is live.

  • On your site, needs work

    At least 4 pieces of content a month

    We publish at least 4 pieces a month, each signed by a named person on your staff, answering one real customer question with a number from your own jobs.

  • On your site, needs work

    Structured data

    We build one clean schema graph: one business node every page references, with your services, prices, service areas, and ratings marked up.

  • On your site, needs work

    Consistent name, address, and phone

    We make your name, address, and phone the same on the site, in the schema, and on every directory listing.

  • On your site, needs work

    AI crawler access

    We open the site to ChatGPT, Claude, Perplexity, and Gemini's crawlers and keep it open.

  • On your site, needs work

    Google Business Profile management

    We manage your Google Business Profile: categories, services, products with prices, photos from your jobs, and weekly posts.

  • Included every month

    Google review replies

    We write a reply to every new Google review; you approve it, then we post it. Ongoing: every new review gets a reply.

  • Included every month

    Local search rank tracking

    We track where you rank on Google Maps for your main services in each city you serve, every week. A crawl of your site cannot measure local rankings; tracking starts on day one.

  • Included every month

    AI Citation Defense

    We run AI Citation Defense: every week we ask ChatGPT, Gemini, Perplexity, and Claude your customers' questions, see who they cite instead of you and from which sources, and do the work that answers it; every month we re-check your data and AI crawler access for drift. Ongoing: every week, for as long as you are a client. A crawl of your site cannot see what the engines say about you; the weekly measurement can.

  • Included every month

    Your ChatGPT business listing

    We set up your ChatGPT business listing with you (you claim it and add CiteMy.ai as your agency), keep its website, description, category, and photos accurate, and add its mentions and website clicks to your monthly summary. A crawl of your site cannot see whether your ChatGPT listing is claimed; most owners have not claimed theirs yet.

  • Included every month

    Job videos

    We make a short narrated video of each finished job from the job's own records (what was done, the materials and options chosen, how long it took, and what the customer said afterward), never with a name or an address, and post it to your YouTube channel and the job's project page with its transcript. Videos start once your job software (Housecall Pro or Jobber) is connected; a crawl of your site cannot see it.

  • Included every month

    Quick-facts page

    We publish a short, dated quick-facts page of questions and answers about your business (license, service area, warranty, what sets you apart, and what AI assistants get wrong), with FAQPage schema and a link from llms.txt; you approve it before it goes live. Every client gets one: it is where the facts AI assistants get wrong are corrected.

  • Included every month

    Distribution where AI engines read

    We publish your data where AI engines read: your sitemap and llms.txt, IndexNow and Bing Webmaster Tools, open data portals and dataset search, structured directories, and Wikidata where you are eligible, each linking back to your site. Ongoing: every new page and dataset is sent out as it publishes.

See everything the monthly service includes

What the new data looks like

Three of the files from steps one to three, for the example business: its llms.txt, its business schema with a price range, and a dataset of its own prices.

exampleroofing.example/llms.txt

# Example Roofing Company

> Tile, shingle, and foam roof repair and replacement in Mesa, Gilbert, Chandler, and Tempe, Arizona. Prices, service areas, and warranty terms below are published as data.

## Services
- [Tile roof underlayment replacement](https://exampleroofing.example/tile-underlayment): price per square foot, how long it takes, the warranty
- [Foam roof recoating](https://exampleroofing.example/foam-recoat): price per square foot and how often to recoat
- [Roof leak repair](https://exampleroofing.example/leak-repair): what a repair costs by roof type

## Data
- [Roof repair prices by roof type (CSV)](https://exampleroofing.example/data/repair-prices.csv): refreshed monthly
- [Service areas by city and ZIP (JSON)](https://exampleroofing.example/data/service-areas.json)

## Answers
- [How long does a tile underlayment replacement take in Mesa?](https://exampleroofing.example/answers/tile-underlayment-time)
- [What does a roof warranty in Arizona cover?](https://exampleroofing.example/answers/roof-warranty)

A plain-text map of the business for AI engines: services, price data, datasets, and answers, each linked. Most sites have none.

Example for Example Roofing Company, a fictional business. Values shortened where marked with an ellipsis.

Example report. The business is fictional; the findings are typical of what our audits find. Results describe what was measured, not a promise of ranking.

Get this audit for your business.

Start with your website. Our AI business engine crawls it the way AI crawlers do and emails you the score, the three findings that matter most, and your implementation plan, in minutes. Free, no call needed.

Free. We crawl your site the way AI crawlers do and email you the score, the three findings that matter most, and a step-by-step plan. We use your email to send them, a few short follow-ups about them (one click stops them), and to reply. By submitting you agree to the Terms and Privacy Policy.