Methodology
How we measure your visibility — and what we can't measure
Every score, every AI answer and every number we quote comes with its method. Here's how the Found, Understood and Chosen scores are built, how we put buyer questions to AI assistants, and where our measurements stop.
Last updated
Principles
What every score rests on
Evidence before scores
Every score is a total built up from specific checks: your website, your Google Business Profile, where you show in Google Maps and how often AI assistants name you. You can always see the evidence behind it: the page or listing, what we found and why it matters. No black box.
Scores are banded, not falsely precise
Each pillar is 0–100 and grouped into Critical, Weak, OK, or Strong. We don't pretend a site is exactly '73.4% optimized'. The band tells you where to focus.
Versioned rules, dated measurements
Scoring weights are versioned, and each audit stores the version it was scored with. After a big change to what the scores count, your account compares only scores made under the new rules. Website checks give the same result for the same site within a couple of points. Your Google listing, Maps place and AI answers can change between runs, so each one carries the date it was measured.
Limits are stated
Website checks are repeatable: the same page gives the same result. Judgment calls, such as how clearly a page reads, come with the evidence they rest on. When an AI assistant's answer could not be used, or a question was not put to it, the report says so and leaves it out of the counts. We never present an estimate as a hard fact.
No invented history
Your first audit is the baseline. We never make up past data or a trend line. Your history starts the day we first measure it.
The three pillars
What each pillar measures
Found — Google & Maps (local SEO)
- Crawlability, indexability, sitemaps & robots
- Titles, headings, and on-page structure
- Local signals: consistent name, address & phone (NAP), service areas & Google Business Profile
- Where you show in Google Maps for your trade and city
- Page speed and core technical health
- Schema / structured data coverage
Understood — AI search
- Your business facts: who you are, what you do, where
- Explicit services & service-area content
- Pages that answer buyer questions directly
- Schema that matches what's on the page
- AI crawlers can reach your pages
- How often AI assistants name you for your buyers' questions
Chosen — a website that turns visits into calls
- 5-second clarity: what you do, for whom, where
- An obvious primary action (call, form, booking)
- Trust signals: your Google rating and reviews next to competitors, proof, credentials
- Mobile conversion & form friction
- Pricing / quote path clarity
How we run buyer questions through AI assistants
We write the questions your customers actually ask ("who's the best foundation repair company in Plano?") and put each one to the assistants through their developer APIs with web search on, on a set date. Reports and monthly plans use the AI assistants we test (currently ChatGPT, Claude and Perplexity).
The consumer apps can answer differently, and answers vary from run to run. So we keep the question set fixed and date every run. In our research, we repeat runs on separate days.
We record whether you're named, whether the answer links to your site, where you appear next to competitors, and who is named instead. An answer we can't use, such as one that was cut off or written after every web search it tried failed, is left out of the counts rather than counted as not naming you. Results are dated, and how often you're named counts toward your Understood score.
AI visibility: readiness and measured mentions
The Understood score has two parts. Readiness is computed from your own site: the structure, clarity and signals AI assistants rely on to understand and describe a business. The same site gives the same readiness result.
Mentions come from the dated buyer questions above: how often the assistants name you, where you're placed when they do, and whether they link to your site. Your account also shows your share of all the business names the answers gave (often called share of voice), but it doesn't count toward the score. Being named counts for more than readiness, because it's the result you're after. Answers vary from run to run, so this part can move between runs even when nothing on your side changed. That's why every run is dated and the question set stays fixed.
Your Google listing and Maps place, checked and dated
When we've checked your listing, it counts toward your Found score: whether it's open, its main category, whether its website and phone match your site, hours, photos, how many reviews you have and how recent they are. Google shows us up to five of your reviews, picked by relevance, not date. If you have more than that, a recent one among them can raise your score, but an old one never lowers it. Where you show in a Google Maps search for your trade and city counts most. Your rating and reviews next to the competitors we compare count toward Chosen.
Listing data comes from Google and is labeled with the date it was checked. The Maps place comes from one search for your trade and city on that date; a buyer's own results also depend on where they search from. A part with no data is left out and the others re-weighted, but a listing we looked for and couldn't find counts as zero.
What your monthly report shows
- What we shipped against your plan's list
- That month's buyer-question results next to the competitors you named
- Your Found / Understood / Chosen scores
- What we'll work on next
In month three, a written before-and-after comparison: the same buyer questions to the same AI assistants, and the same site re-audited with the same scoring version.
See a sample monthly reportHow we check the numbers we quote
Every outside number on this site comes from one registry of verified statistics. Each one keeps its source's own wording and sits next to the source, the publication date and a link, so you can check it yourself. We re-check each one against its primary source on a set schedule.
Facts about Texas markets follow the same rule: we show one only after checking it against its primary source, and we label our own analysis as ours.
Most of the local-search numbers we quote come from BrightLocal, which sells local SEO software. We use them because they're the largest recent US surveys on local search and AI, and we show independent sources next to them where they exist.
How we count our own audit data
When we publish figures from our own audits, they come from the audits we run on Texas home service websites. Each business counts once, using its first audit, run before we do any work for it. For AI answers, we ask each assistant buyer questions about the business's trade and city, such as "best foundation repair in Houston, TX". We ask through the assistant's developer API with web search on, and record whether the business is named. An answer we can't use, such as one that was cut off or written after every web search it tried failed, is left out of the counts.
What these numbers are not. They are not a random sample of Texas contractors. We usually audit a business because we think its website can be improved, so these figures likely describe weaker-than-average sites. API answers can differ from what the consumer apps show, and answers change from day to day. We report each assistant separately, publish the sample size and dates next to every figure, and never show a figure based on fewer than 30 businesses.
How we report client results
When we publish client results, each one goes up client by client, with that client's permission, next to a count of every client measured the same way, including any that didn't improve. We show a median and a range only once 10 clients reach day 90, and a single headline number only at 20. The measure is the change, in percentage points, in how often AI assistants name you (the median of 3 runs), compared with the competitors you named. We record every client's starting point on day 0.
Questions about how we measure
Why do AI answers change from one run to the next?
Each answer is generated fresh, and the web results, models and sources behind it keep changing. So we keep the question set fixed, date every run, and report how often you were named rather than one answer. In our research, we repeat each run on separate days.
Is my AI score the same as being mentioned?
Not quite. Your Understood score combines how ready your website is for AI (whether it says clearly who you are, what you do and where, in a form machines can read) with how often AI assistants actually name you when we ask them buyer questions on a set date. Being named counts for more. Each report shows both parts, the questions we asked and who was named instead.
What we won't do
Some tactics get quick attention and cost you later. We don't use them, for you or for ourselves.
- No mass-produced AI articles or copy-paste city pages. We publish fewer pages, built from your real jobs and your customers' real questions.
- No fake reviews, no review gating and no paid-for reviews. We ask every customer the same way, not only the happy ones, and never trade a discount for a review.
- No fake posts, fake offices or fake profiles. Not on Reddit or Quora, not on Google Maps. When we take part in a discussion, we say who we are.
- No bought links, no parasite pages and no paid "best of" spots. That includes self-ranked "best agency" lists. Google's spam policies cover all of these.
- No charging for work we miss. Every item in your plan, except the as-needed ones, has a credit value. If we miss one, its credit comes off your next invoice, or, if it was your last month, we refund it.
- No hidden prices and no fake urgency. Every plan price is on our pricing page.
See where your business stands
We review your website and Google Business Profile by hand and email directional Found, Understood and Chosen scores with your first fixes.