Real checks, explained

What two real AI visibility checks found

We checked two real businesses using the same process available to every Ghosted visitor: 10 buying questions asked to ChatGPT and Claude with web search on.

The businesses are described by industry and market because they did not ask to be publicly identified. These are dated snapshots, and results may change when the questions are asked again.

See exactly how the check works

A speciality coffee roaster with shops across the UK

Tested:: August 2026
Market:: United Kingdom
Method:: 10 questions, ChatGPT and Claude, web search on
Status:: Awaiting publication and re-check

The roaster appeared in 2 of 20 answers. It was absent from the other 18.

Business mentionedBusiness not mentioned

Visibility range0–20

A range is shown because assistants can disagree and their answers can vary.

The problem

The roaster had an established real-world business but appeared in only 2 of 20 answers about where to buy speciality coffee.

What the check found

Its robots.txt restricted important AI crawlers. Key business information was also difficult for assistants to access and quote directly from the website.

Files Ghosted prepared

Finished files the business can publish directly or send to its web developer.

Updated crawler-access instructions
A robots.txt that lets the assistants' crawlers read the site.
Google business verification recommendations
The listing details assistants lean on for local answers.
Clearer business and location information
Who the business is and where it operates, stated on the page.
Quotable category-positioning copy
Two lines written to be repeated back by an assistant.
Developer handover note
Where each file goes and what to change, for whoever runs the site.

Next step

Today

0–20

Target once published

70–75

Projected, not measured. Today no assistant names this business in 8 of the 10 questions. If the prepared files earn it a single mention in each of those, the same scoring puts it at 70 to 75. We will ask the same 10 questions again after publication and show what actually happened.

A pre-owned luxury retailer selling across Europe

Tested:: August 2026
Market:: Europe
Method:: Earlier method, re-check required
Status:: Awaiting publication and re-check

Current-method re-check required

This business was originally tested using an earlier version of the Ghosted method. We are rerunning it using the current 10-question, two-assistant check before publishing comparable figures.

The problem

When asked where to buy authenticated pre-owned designer bags, the assistants confidently recommended larger marketplaces. This retailer was absent from most answers.

What the check found

The website did not clearly and consistently describe the business in language the assistants could reuse. Important information about authentication, geographic coverage and pricing was also difficult to identify.

Files Ghosted prepared

Finished files the business can publish directly or send to its web developer.

AI crawler access recommendations
Which crawlers can read the site today, and what to change.
Clear business-definition copy
What the business is, in language an assistant can reuse.
Structured data describing the business
The same facts in a format machines read directly.
Copy addressing common pricing questions
The price questions customers ask, answered on the page.
Developer handover note
Where each file goes and what to change, for whoever runs the site.

Next step

The files have been prepared. This business is being rerun on the current 10-question check, and its figures and target will appear here once that run completes.

What would the assistants say about your business?

Ask the same 10 buying questions for your own market. See who gets recommended, which sources shape the answers and what may be missing from your website.

No signup. No card. Full findings in 2–4 minutes.