Example report · fictional firm
This is the complete report AgentSignal produces, run on Vantage Accident Attorneys, a fictional personal injury firm in Chandler, Arizona. The firm, its competitors, the answers and their sources are invented; every number below was computed from them by the same pipeline that measures a real firm, and the scan history before the latest scan is illustrative.
Fictional firm and competitors · No live AI systems queried
AI VISIBILITY REPORT · Pro run
Chandler, ArizonaPersonal injury law firmMeasured Aug 19, 2026
Getting found is your biggest opportunity.
AI Discovery
0113 of 30 answers mentioned you
How often AI mentions your firm.
How often AI names your firm in the completed test answers.
Answers naming your firm ÷ completed test answers × 100. Rounded to a whole percentage from 13 of 30 completed test answers.
AI Selection
02When mentioned, 2 of 13 recommended you first
When AI mentions you, how often does it recommend you first?
Of the test answers that name your firm, how often AI recommends it first.
First recommendations ÷ test answers naming your firm × 100. Rounded to a whole percentage from 2 of 13 test answers naming your firm.
ONE FUNNEL, ONE DENOMINATOR
30 planned · 30 completed · 0 failed and excluded · 2 diagnostic reruns not counted. Each rung is the one before it minus the answers that left there, and the partition reconciles.
Recommended first overall2 of 30 (7%)
How often AI recommends your firm first across all completed test answers, not only the answers that name it. First recommendations ÷ completed test answers × 100. Rounded to a whole percentage from 2 of 30 completed test answers. Of 30 test answers, 13 named your firm and 2 of those recommended it first.
90% recommended another firm first. 3% gave no single first recommendation.
Personal injury law firm · 30 client situations × 1 AI model: 30 planned test answers; 30 completed. A test answer is one model’s response to one situation, not a person and not an enquiry. Every count on this page is over the 30 test answers that completed; the ones that did not are excluded rather than counted as a result. Results describe the tested scope, not the entire firm.
Your firm was named in 13 answers and was not mentioned in 17.
Your firm was absent from 57% of the completed test answers. Start by making your relevant services easier to find and verify.
The 12 recommended changes in this report, grouped into the 3 things to work on first. Open a fix for the steps and the evidence.
Not named in 4 of 4 car accident test answers
One page: Car Accident · One page about Slip And Fall · One section: Truck Accident · One page about Aviation And Boating Accidents · One page about Catastrophic Injury · One page about Birth Injury · One page about Bicycle Accident · One section about Motorcycle Accident · One page about Burn Injury
Contributed to 17 test answers
One third-party profile, claimed and completed
Recommended first in 0 of 6 burn injury, defamation, hit and run, mass tort, negligent security, nursing home abuse test answers
Evidence published on a page, with its source
Not named in 2 of 6 defamation, hit and run, rideshare accident, slip and fall test answers
One page: /attorneys
Reach can overlap between changes — do not add these counts together. Nothing here predicts a result: a re-test compares matched observations, so a later benchmark showing a different outcome is a measured difference between two sets of answers, not proof an edit caused it. A threshold like “by at least 2” is a fixed reporting rule: at least two decisions must move, or one when the re-tested set holds three or fewer. It is not derived from this sample's variability, so it does not establish that a change of that size is more than ordinary run-to-run movement. It is a reporting threshold, not a target and not a prediction.
sample size × stage of loss × fixability. Competitors' published results are being cited and yours are not — ordered first: it rests on 6 of 30 decisions, they were lost after the firm was surfaced in 8 of the 8 it was observed in, and it is a medium-effort change to a page you already have. Rule: sample size × stage of loss × fixability.
Priorities are grouped by what the diagnosis recorded about each change — the cause behind it and the kind of artifact it makes — and ordered by whether anything in the group blocks the rest, then by the recorded impact, then by the plan’s own printed position. Every change keeps its own counts inside its fix.
Low effort A change to pages or settings you already have.Medium effort Writing or reworking pages, or updating your profiles on other sites.High effort Ongoing work, such as earning reviews or publishing original material.A change carries the lowest effort among the findings it answers. The level describes the kind of work, not how long it takes or what it will change.
By case type, largest samples first. Every figure is a count over the answers tested for that row — a small sample is a starting point, not a pattern.
| Case type | Mentioned in | Recommended first in | Status |
|---|---|---|---|
| 4 answers tested | 0 of 4 | 0 of 4 | Not mentionedAI never surfaced you in any of the 4 car accident test answers it ran. Selection: not measured — no answer here named your firm. |
| 3 answers tested | 1 of 3 | 0 of 3 | Never firstAI never surfaced you in 2 of the 3 slip and fall test answers it ran. Selection: 0%. Recommended first among the answers naming your firm — 0 of 1. |
| 3 answers tested | 0 of 3 | 0 of 3 | Not mentionedAI never surfaced you in any of the 3 truck accident test answers it ran. Selection: not measured — no answer here named your firm. |
| 1 answer tested | 0 of 1 | 0 of 1 | Not mentionedAI never surfaced you in any of the 1 aviation and boating accidents test answer it ran. Selection: not measured — no answer here named your firm. |
| 1 answer tested | 1 of 1 | 0 of 1 | Never firstAI surfaced you in 1 of 1 bicycle accident test answer, and did not recommend you first in 1 of those. Selection: 0%. Recommended first among the answers naming your firm — 0 of 1. |
| 1 answer tested | 0 of 1 | 0 of 1 | Not mentionedAI never surfaced you in any of the 1 birth injury test answer it ran. Selection: not measured — no answer here named your firm. |
Ordered by how many answers were tested for each case type — never by an estimate of demand, case value or revenue, none of which this scan measures. Open a case type to read the recorded answers behind its counts.
Every firm AI named in these 30 test answers, and how many of them recommended each one first.
Preview: 5 of 7 rows — the 5 firms recommended first most often. The scale and every count are over all 7 firms on this list. Bars are counts of answers, on one scale running to 8. Open a firm to read the recorded answers it was recommended first in, and the reason the assistant gave.
WHAT THE WINNING ANSWERS HAD
Counted from what the assistants said while recommending another firm. Nothing here was read from a competitor’s website — this scan crawls your site and nobody else’s.
EXAMPLE TEST SITUATIONS
Your firm was not mentioned
A fall on a friend's icy steps (Trust-Sensitive Client)
WHAT THE CLIENT ASKED“I broke my ankle when I slipped on the icy front steps of a friend's house and I do not know whether that is even a case when it is someone I know. I am in Chandler, Arizona. I…”
WHY IT RECOMMENDED STONEGATE LAW OFFICES, IN ITS OWN WORDS“A personal injury page that turns up easily in search, and it is the closest match to your constraints”Run 00000000-0000-4000-8000-000000000001
Your firm was named, but not recommended first
Cyclist struck by a turning driver (Local-Firm Client)
WHAT THE CLIENT ASKED“A driver turned right across me while I was riding my bicycle to work and I came off hard — I have a broken wrist and a fractured collarbone. I am in Chandler, Arizona. I was in…”
WHY IT RECOMMENDED REDROCK LEGAL GROUP, IN ITS OWN WORDS“Offices across Arizona and the resources to fund experts against an insurer, and a large team and a name most people here recognize”Run 00000000-0000-4000-8000-000000000011
Your firm was recommended first
Serious injury at work with a third party involved (Local-Firm Client)
WHAT THE CLIENT ASKED“I was injured on a construction site when equipment operated by a subcontractor came loose and crushed my foot. I am in Chandler, Arizona. My employer has put in a workers comp…”
One record per outcome, taken in the order the log stores them. They show what each outcome looked like; they are not a sample and they do not describe how often each happened — the counts above do that. A quoted sentence is what the model said in that one answer, which establishes the outcome and not the reason behind it.
What the diagnosis attributed the lost decisions to. Open a cause for what it means.
AI is not finding you
In the affected client journeys your firm never came up at all, so nothing else about how you present yourself had a chance to matter.
17 of 28 decisions not won.
Attributed by the diagnosis rules over 17 recorded answers: the answers are the observation, the cause is the inference.
Nothing independent vouches for you
Nothing outside your own website supported the recommendation — no results, no recognition, no reviews an assistant could quote back.
6 of 28 decisions not won.
Attributed by the diagnosis rules over 6 recorded answers: the answers are the observation, the cause is the inference.
The people who would do the work cannot be checked
The client was deciding on who would actually handle the matter, and nothing about a named person was retrievable for anyone at your firm.
3 of 28 decisions not won.
Attributed by the diagnosis rules over 3 recorded answers: the answers are the observation, the cause is the inference.
You read as a generalist
Assistants could tell what area you practise but not that you handle this particular kind of matter, so they recommended a firm that looked specific.
2 of 28 decisions not won.
Attributed by the diagnosis rules over 2 recorded answers: the answers are the observation, the cause is the inference.
Counts are over 28 decisions not won — another firm was recommended first, or no firm was. A cause explains why a decision was recorded the way it was; it is not a prediction that removing it wins the decision back.
Read from 5 pages of your own site. These are eligibility checks, not causes of the results above.
5 of 5 documented crawlers have no blocking rule.
Nothing in 5 pages read stops an assistant reading or indexing them.
1 issue weakens what an assistant finds when it does read you.
1 observation recorded, none of them blocking.
FOUR DIFFERENT QUESTIONS
Crawl permissionMeasured in this scan
What does your robots.txt allow a named crawler to request?
5 of 5 documented crawlers have no blocking rule.
Read from your robots.txt, against the crawler tokens each vendor documents for itself.
Does not establish: That any crawler requested a page, retrieved one, or was able to read it. A permission is a statement about what is allowed, not about what happened.
Observed retrievalMeasured in this scan
Did an assistant actually read your site while answering?
Your own domain was cited in 13 of 30 completed decisions (13 citations).
Citations the assistants themselves returned. Separately, our own crawler fetched 5 pages of your site — that is us reading you, not an assistant.
Does not establish: That an uncited page was unreadable. An assistant that answered without citing you may still have read you, and a citation is evidence of reading rather than a measure of it.
IndexingNot measured
Are your pages in a search index?
Not checked in this scan.
Nothing in a scan queries a search index. A page declaring `noindex` is recorded as a blocker above, but the absence of that declaration is not a verified index entry.
Does not establish: Anything at all — in either direction. Treat it as unknown rather than as a pass.
Mentions and recommendationsMeasured in this scan
What did the assistants actually say about your firm?
Named in 13 of 30 completed decisions; recommended first in 2.
The benchmark's own answers, counted over completed primary decisions.
Does not establish: A cause. These counts say what happened, not why, and nothing above them was shown to have produced them.
AI ACCESS RULES, CRAWLER BY CRAWLER
5 of 5 have no blocking rule. What each vendor documents for itself:
Why it is checked: A retrievable page for each kind of matter the firm genuinely handles
WHAT TO DO
General steps for this family of checks, not written for this issue. Read them against what we actually read below.
Not if: The firm does not actually handle the matter type — never invent a practice.
WHAT WE READ
The page at https://vantageaccidentlaw.com/ has 32 words of visible text.
200 characters after navigation and scripts were stripped, against a 150-word floor for a key page.
The page at /attorneys has 24 words of visible text.
143 characters after navigation and scripts were stripped, against a 150-word floor for a key page.
The page at /practice-areas has 103 words of visible text.
645 characters after navigation and scripts were stripped, against a 150-word floor for a key page.
The page at /practice-areas/car-accidents has 103 words of visible text.
601 characters after navigation and scripts were stripped, against a 150-word floor for a key page.
What we do not claim: That firms with dedicated pages are recommended first because of them. We observed that firms surfaced for these matters were credited with specific work; the page makes that credit possible for this firm.
Why it is checked: Basic page hygiene: titles, headings, HTTPS, redirects
No <meta name="description"> was found in the served HTML.
WHAT TO DO
Write a description for these 5 pages
WHAT WE READ
The page at https://vantageaccidentlaw.com/ has no meta description.
No <meta name="description"> was found in the served HTML.
The page at /attorneys has no meta description.
No <meta name="description"> was found in the served HTML.
The page at /free-consultation has no meta description.
No <meta name="description"> was found in the served HTML.
The page at /practice-areas has no meta description.
No <meta name="description"> was found in the served HTML.
The page at /practice-areas/car-accidents has no meta description.
No <meta name="description"> was found in the served HTML.
WHAT THIS DOES NOT ESTABLISH
The same client situations, measured again.
Fictional example: the earlier scans are illustrative; the latest is the computed result.
Re-test after the changed pages have been re-crawled and indexed. Google documents that crawling can take anywhere from a few days to a few weeks (Google Search Central, Ask Google to recrawl your URLs, updated 2025-12-10).
AI answers change, and a before/after difference does not prove causation or predict enquiries. Work your team recorded as implemented, a check a later crawl could confirm, and a different benchmark outcome are three separate things.
Each measure below opens the answers it was counted from. Nothing here is a summary of them — they are the records.
30 client situations × 1 AI model: 30 planned test answers; 30 completed. A test answer is one model’s response to one client situation, not a person and not an enquiry. Failed attempts are excluded from every count rather than counted as a result.
Each AI model was asked through its own API. That is not a test of the vendor’s consumer app, which may retrieve and order its results differently. AI answers vary between runs, and a difference between two scans is a measured difference between two sets of answers — never proof that an edit caused it, and never a forecast of enquiries.
A threshold like “by at least 2” is a fixed reporting rule: at least two decisions must move, or one when the re-tested set holds three or fewer. It is not derived from this sample's variability, so it does not establish that a change of that size is more than ordinary run-to-run movement. It is a reporting threshold, not a target and not a prediction.
TRY IT FOR YOUR FIRM
Free first scan · No credit card
Next
Example report: there is no live scan to quote.
Re-test after publishing