# Treat every public file as a claim. Then check it.

Flowpane fetches the files that govern your site, scores each one, checks them against each other and against what crawlers and Google actually see, and keeps a record you can defend.

## The loop

1. **Add.** One URL, a pasted list, or a CSV. Platform and audience are detected, and kept apart from what you declare.
2. **Verify.** A file, a DNS TXT record, or a meta tag. One value covers the whole workspace.
3. **Check.** Fetch, parse, assess, score, store. Every run is immutable.
4. **Review.** Findings with the exact line, ranked by severity: Issue, Warning, Recommendation, Information.
5. **Publish.** You publish the fix through your own host or CMS. Flowpane proposes, you decide.
6. **Monitor.** Checks repeat on your plan's schedule, from weekly to hourly. The next public check moves the score.

## One number, built from eight, weighted by what they cost you.

Each file gets its own score from 0 to 100. The site score weights them by what they cost you, and skips what does not apply.

- Sitemaps: weight 35 (required)
- robots.txt: weight 15 (required)
- security.txt: weight 12 (required)
- Cookies and consent: weight 12 (required)
- ads.txt: weight 8 (conditional)
- llms.txt: weight 7 (recommended)
- ai.txt: weight 6 (recommended)
- humans.txt: weight 5 (recommended)

- **Only what applies to you.** robots.txt, sitemaps and security.txt always count. ads.txt only counts if you sell advertising, so a site without ads is never marked down for it.
- **The last good result.** The score uses each file's last good result, so a failed fetch does not wipe out good evidence.
- **Blocked is not missing.** If a firewall challenges the crawler, the result is Blocked and the site is shown as unscorable. You never get a flattering partial score.

Bands: 80 to 100 good, 60 to 79 needs attention, 0 to 59 failing.

## Sixteen rules for files that contradict each other.

Coherence is reported beside the site score, so a set of individually valid files that disagree still shows up.

- robots.txt blocks URLs the sitemap asks search engines to index
- The sitemap lists pages marked noindex
- A sitemap exists but robots.txt never declares it
- llms.txt links to URLs robots.txt blocks
- robots.txt and ai.txt disagree about AI training
- ai.txt and security.txt give different contacts

## What you declared, what is reachable, what Google kept.

Three sources, kept apart on purpose. Flowpane never blends them into one number, because the gaps between them are the findings.

- **Public pull** (What is on the wire?) Every artefact, fetched over public HTTPS from the canonical origin by a crawler that identifies and signs itself.
- **Deep Scan crawl** (What is reachable?) A breadth-first crawl from the homepage that respects robots.txt and Crawl-delay, then compares every page with what the sitemap declares.
- **Search Console** (What did Google accept?) Read-only. Submitted sitemaps, counts, and a bounded URL inspection, shown beside Flowpane's own observation with timestamps.

## Fix it without guessing, and without inventing anything.

Every failing or missing file opens in a workbench with the current file beside a proposed one.

- **Fix and Fix all.** Deterministic, safe patches for mechanical problems. No model involved.
- **AI Assist.** Drafts text for a missing or failing file. The draft must pass the same factual-safety and validation checks as anything you write, and a rejected draft does not use your allowance.
- **Safe starters.** Missing files start from a conservative template that never invents a contact, a policy, a publisher ID, or an AI stance.
- **Review, then publish.** Validate, then on Agency approve an exact saved revision. You publish it. A draft never changes a score.

## Know exactly what you are telling AI systems.

robots.txt is assessed against a directory of named AI crawlers, from GPTBot to ClaudeBot to Google-Extended. Content-Signal directives are read, and ai.txt and llms.txt are checked against both. The result is a grade from A to F with the checks behind it.

## A record you can send, and a queue you can work.

- **Governance reports.** A PDF with an executive summary, per-file state, cross-file coherence, recommendations, and a section on what the evidence can and cannot prove. From Starter.
- **Workspace analytics.** Needs-attention ranking, file coverage and AI readiness across every site in a workspace, with CSV export from Pro.
- **Roles per workspace.** Viewer, Editor, Reviewer, Publisher and Admin. Workspaces map to clients, brands or teams.
- **White-label.** On Agency and Enterprise, reports carry your logo and can drop the Flowpane footer.

## Clear edges make it easier to trust.

- **Not an on-page SEO plugin.** Plugins shape individual pages. Flowpane checks the site-level files around them. Use both.
- **Not an uptime monitor.** It does not page you when the site is down. It tells you when the site's public claims are wrong.
- **Not a compliance certificate.** Scores are product models. Reports state their evidence boundaries rather than promise legal outcomes.
- **Not a tool that edits your site.** Flowpane reads your origin. You decide what gets published, and when.

Join the beta: https://flowpane.com/beta
Documentation: https://docs.flowpane.com
Crawler identity: https://crawler.flowpane.com