How it works
Mechanics first, then prompts, then fixes.
An AI answer is assembled from pages a crawler could fetch, read, and justify citing. So we start with whether crawlers can reach you, then check whether you're named, then tell you what to fix.
Our position, stated plainly
We can't see inside any model's ranking, and neither can anyone selling you a tool. What we can do is check the inputs, meaning whether crawlers can fetch and read your pages, and the outputs, meaning whether you're named for the questions your buyers ask.
So we report evidence: status codes, robots rules, missing files, broken URLs, and cited-or-missed per question. Every task links back to the finding that produced it.
Step 1. Get invited and sign in
Visibility Copilot is in private beta. Request an invite; once you're in, sign in with Google. Every account gets a workspace you can share with teammates.
Step 2. Run your first report
Enter the domain you want analysed and pay $2 plus taxes for one full report. The trial is limited to one per email and one per domain.
- Site audit. Runs right away: about 40 pass/fail checks on crawl access, indexability, sitemap and llms.txt, readable HTML, structured data, and on-page basics.
- Pick your prompts. We suggest up to 10 questions based on your domain. You edit them, then run the prompt check.
- Prompt results. Each question is marked cited or missed, broken down by engine, with the top rival, sources, and suggested actions.
- Tasks. Failed checks and lost prompts become a ranked task list, each with what we found, what to do, and a draft where one helps.
Step 3. Weekly monitoring on a plan
If the report is useful, pick a plan. Each monitored domain is then re-scanned once a week, on the same weekday as its first report.
- The site audit runs again, which catches regressions like a new firewall rule or a broken sitemap entry
- Your prompts run again, so you see movement rather than one snapshot
- Tasks that now pass close themselves; new findings open new tasks
- Home shows this week against last week
Shipped a fix mid-week? Re-run the site audit from the report page and see the result right away. Prompt results stay on the weekly schedule.
Progress looks like more prompts where you're named, more checks passing, and fewer open tasks. You can count those. We'd rather show you those than a single index moving from 58 to 61 for reasons we can't defend.
What the site audit checks, and why
- Crawl access. robots.txt rules for Googlebot, Bingbot, OAI-SearchBot, plus AI crawlers such as GPTBot, ClaudeBot, PerplexityBot. We also make one live request with an AI-crawler user agent, because a firewall can block crawlers even when robots.txt says Allow.
- Indexability. HTTPS, redirects, redirect chains, canonical tags, and
stray
noindex. Classic technical SEO that still decides whether a page can be cited. - Discoverability. Whether
sitemap.xmlandllms.txtexist, and whether a sample of the URLs inside them load. - Readable HTML. Whether your copy is in the raw HTML. Most crawlers don't run your JavaScript, so a client-rendered page can look empty.
- Structured data. Whether JSON-LD is present and names your organization or product, plus Open Graph tags.
- On-page basics. Titles, meta descriptions, a single H1, image alt text, security headers, and a sample of internal links.
The audit is deliberately shallow: the homepage, robots.txt, the sitemap, llms.txt, and a small sample of URLs. It catches the blockers that matter most. It isn't a full-site crawler and doesn't replace one.
How a finding becomes a task
Examples of what lands in your list:
- Publish
/llms.txt, with a draft to paste and edit - Remove the
Disallowfor the AI crawlers named in the finding, with a robots.txt draft - Fix or unlink two internal URLs returning 404 and 410, each linked so you can open it
- Add the missing security headers listed in the finding
- Publish an answer-first page for a question where a rival is named instead of you
Ship it, mark it done, and let the next scan confirm it.
What we won't do
No auto-sent outreach. No posting on your behalf. No daily notification stream. No pages published to your site automatically. And no single proprietary score that claims to explain AI visibility. If we can't show you the check behind a number, we won't charge you for it.
See which questions you lose, and why
We're in private beta and onboarding a small number of teams. Once you're invited, your first report costs $2 plus taxes, and you only subscribe if it's useful.