The outside watch for your firm · AI answers about your firm

When AI sends your client to the wrong number

Your clients now ask an AI assistant for your phone number or your client login. The answer comes from pages your firm does not control. A wrong one sends the client to whoever answers. We ask the same questions first, check each answer against your own website, and show you what your clients are told.

Attackers now use AI to read a firm from outside the way a search engine does: the website, directories, search results and what AI assistants say. We read it first, and every finding is human-verified before your assessment is released.

Check my firm What we check

Read-only, from outsideNothing tested or logged intoHuman-verified

What we check for thisRead-only

Phone number

Whether the number an assistant gives is one your own website lists

Client login

Whether the sign-in page it gives is on your domain, or on a provider your site links to

Payment and wire

In the assessment: whether it gives payment details or a payment page that is not yours

Where it came from

In the assessment: the page the answer was taken from, opened and quoted

Human-verified before release. Every finding carries its proof. Read-only, from outside.

01 · What we check

Four answers that send a client to the wrong door.

On 29 September, SecurityWeek reported Users searching for ChatGPT: the way in was an AI answer pointing people to the wrong place.

A phone number your site does not list

A client asks for your number and is given one that appears nowhere on your website. It may be an old line, a directory’s number or somebody else’s. We show you the number and the answer it came in.

A login page that is not yours

A client asks where to sign in and is sent to a page off your domain. A sign-in page you do not control can collect your client’s password.

Payment details that are not yours

In the assessment we also ask how to pay you and where to send a wire. An answer that gives bank details, or a payment page off your domain, is shown to you word for word.

A web address nobody has registered

Sometimes an assistant gives a web address for your firm that no one owns. We check whether each address it gives is registered, and whether it is yours.

02 · What we read

Proof, not scores. Every finding carries its evidence.

The question and the answer

Each finding carries the exact question, the assistant that answered, the answer and the date, so you can ask the same question yourself. Your free outside reading asks Claude and Gemini two questions your clients ask: your phone number and your client login. The assessment and the watch ask Claude, ChatGPT search, Gemini and Perplexity six questions your clients ask: your phone number, your client login, how to pay you, where to send a wire, which email to use and whether your firm is legitimate, and read your Google business listing for the phone number.

Asked more than once

AI answers change from one ask to the next. In the assessment a wrong answer is asked again, up to three asks in all, and the evidence says how many of them gave the same answer.

Where the answer came from

In the assessment we open the pages the assistant cited and find the one that carries the wrong detail, word for word. That page is what needs correcting.

The correction, prepared

We prepare the correction and re-check until it’s right: the page to edit, with a direct link, the text from your own website to paste, and the same details for your business listings and your professional directory. You or the person who runs your website makes the edit. On monitoring we ask the same question again every week, and a finding is closed only when the right answer comes back in two of three asks.

What it cannot see

What a client asked in private, or which assistant they used. We do not edit your listings and we do not write to the AI companies. A number your site does not list may still be a working line of yours. The finding is what your client is told.

03 · What we never say

A wrong answer is a fact to correct. We say where it came from.

We do not tell you that a wrong answer is fraud, or who put it there. An answer can be wrong because a directory is out of date. What we give you is the question, the answer, the date and the page it came from, and the correction for that page.

How we are different

A different job from the people you already work with.

Your IT provider

Excellent at running your systems, and makes the changes. We read what the outside world sees and hand them the exact fix, with the check that proves each one. Different job, and independent: we are paid by no one who sells the fixes.

A penetration test

Tries to break in and needs access. We are not a pen test: nothing attacked, nothing installed, no access given. We read only what is already public, from outside, and log into nothing.

Your insurer's scan

Feeds your premium and eligibility. We show you the same outside view first, in plain words, so you walk into renewal knowing the answers.

Look-alike domains How we use AI Request your outside reading

For your firm

See what AI assistants tell your clients about your firm, before a client acts on it.

The External Exposure Check: your outside reading, the openings a stranger can see on your firm today, each with its proof. Read-only, from outside.

Check my firm