You do not know you are invisible
Someone asks for the best emergency plumber in Manchester. ChatGPT names three competitors. Your business is not mentioned, and nobody tells you.
crawrix · SEO · AEO · GEO
What does AI say when someone asks for a business like yours?
Find out if models recommend you. Get a ranked plan. Ship fixes when you approve them.
Free scan, no card. Or start Pro and skip the waitlist.
crawrix report4 models · weekly
AI Visibility
24/100
- AI Readiness
- 61
- Overall
- 41
What AI answered
- “Best emergency plumber in Manchester”Not mentioned
- “24 hour boiler repair near me”Competitor named first
- “Gas safe engineer Manchester”Named, 2nd of 3
What AI can read
- GPTBot allowedpassing
- ClaudeBot allowedpassing
- /llms.txtmissing
- LocalBusiness JSON-LDmissing
3 fixes are waiting for your approvalNothing shipped yet
Probed across four models, then averaged
- ChatGPT
- Claude
- Grok
- Perplexity
The shift
Demand routes through AI answers before it hits your site.
If ChatGPT never names you, ranking next quarter will not matter. Keyword dashboards cannot see those answers, and agencies sell hours against them.
Tools list keywords. You still implement
Keyword dashboards do not see ChatGPT, Claude, Perplexity, or Grok. They hand you a list and leave the work on your desk.
Hours billed, work not shipped
An agency runs $300 to $800+ a month on a monthly cadence and still rarely measures AI visibility. A one-off diagnosis is often $500 to $2,500.
There is a fourth fear, and it is the quiet one. That an AI will change the live site while nobody is watching. crawrix cannot. Every write waits for your approval, and it never pushes to main.
How it works
A continuous loop, not a one-off PDF.
Measure what AI says about you. Rank what to fix. Ship only what you approve. Then measure again.
- Measure
- Rank
- Approve
- Re-measure
- 01Measure
Ask AI what customers ask
The same questions people type into ChatGPT, Claude, and Perplexity about your services and your area.
- 02Measure
Check what AI can read on your site
Assistants only recommend businesses they can read and verify. This part is a plain crawl, not a model guess.
- 03Rank
A short ranked plan
Impact first, in owner English. Every item says why it matters, so you can skip the ones you do not care about.
- 04Approve
Fixes ship only when you approve
Nothing touches the site on its own. crawrix never pushes to main.
Then re-measure. Weekly on Pro and Max, daily on Ultra Max. The free scan is a one-off, so you run it when you want it.
Scores
Three numbers, published formulas.
Answers are averaged across models so one coin-flip is not the number. When a provider is missing, the report says so instead of inventing a zero.
AI Visibility0 to 100
When people ask AI the questions your customers ask, do you show up?
Moves in weeksAI Readiness0 to 100
Can machines find, read, and understand you?
Moves in days, once you ship fixesOverallBlend
AI knowledge 35%, AI search visibility 35%, site readiness 30%. A missing provider degrades the report. It never invents a zero.
Follows the other twoThe probe fleet
Four models answer the questions your customers ask. A cheaper model judges the answers.
- ChatGPTopenai/gpt-5
- Claudeanthropic/claude-sonnet-4.6
- Grokxai/grok-3
- Perplexityperplexity/sonar
The readiness crawl
A deterministic crawl, not a paid model. It either finds these or it does not.
- robots.txt rules for GPTBot, ClaudeBot, PerplexityBot, Google-Extended
- /llms.txt
- Organization and LocalBusiness JSON-LD
- sameAs social profiles
- Sitemap
- Answer-ready homepage
- Consistent business name
Technical
Titles, meta, OG, canonical, robots, sitemap, JSON-LD, alt text, lang.
Performance
Core Web Vitals, images, fonts, JavaScript weight.
Content
H1, heading order, copy, internal links.
Crawl
Status codes, noindex, hreflang, orphan pages, AI crawlers.
AEO
Extractable answers, FAQ and HowTo schema that matches the text on the page.
GEO
Being known and cited by generative engines.
These are best-effort diagnostics. They are not ranking or citation guarantees, and no honest tool can sell you those.
Outcomes
Three things you actually get.
Not a bigger dashboard. Proof, a plan, and work that ships on your word.
Know if AI recommends you
Named, absent, or losing to a competitor. Across models, tracked over time instead of guessed once.
Own the pages machines should cite
When a competitor wins the prompt, you get the page to build. Not the advice to blog more.
Ship fixes without losing control
Every write pauses for approval. Nothing lands on main alone.
Compare
An agency, a keyword tool, or crawrix.
Tools list keywords. Agencies sell hours. crawrix measures AI visibility, ranks what to do, and optionally ships the technical fixes you approve.
| Capability | SEO agency | SEO tools | crawrix |
|---|---|---|---|
| AI visibility | Rarely measured | Emerging | Probes, score, history |
| Cadence | Monthly | On demand | Continuous |
| Fixes | Human hours | You implement | Approved PRs |
| Language | Variable | Dashboards | Owner English |
| Core price | $300 to $800+ | Per seat | $40 to $250 |
A good fit
- Service businesses spending $0 to $800 a month on SEO, including first-time buyers.
- Owners who want proof in plain English instead of another dashboard.
- Developers with a GitHub-hosted site, especially Next.js.
- WordPress owners already running Yoast, Rank Math, or AIOSEO. crawrix writes through them.
- Anyone with no repo at all. Email login exists for you.
Not us
- Businesses paying $1,500+ for content and link building.
- Enterprises with procurement and an agency contract already in place.
- Anyone who wants crawrix to write content or buy links.
Pricing
Every paid plan is the whole loop.
Usage is the ladder, not feature locks. USD, monthly, no annual contract, no free trial. When you hit the ceiling, crawrix tells you that you have used this week's fix allowance.
Free Scan
$0one-off
See how AI engines talk about your brand. Top issues, plain English.
- One website
- Full SEO and AI visibility audit
- AI Visibility and AI Readiness scores
- Top 3 to 5 issues
- No card
Pro
Popular$40per month
The full loop for one site. Usage is the limit, not features.
- 1 website, more at $35 a month each
- Weekly tracking
- Technical fixes with approval
- Chat that acts
Max
$120per month
Same product, up to 10 sites, more headroom.
- Up to 10 websites
- Frontier models
- Weekly tracking
- Agent CLI and MCP when they ship
Ultra Max
$250per month
Heavy throughput and daily re-measurement.
- Unlimited websites
- Highest allowance
- Daily re-measurement
Extra location
$15 a monthSame site, second town or branch. More prompts.
Additional domain
$35 a monthA separate website. Raises your site cap by one.
The free scan gives you both scores and your top 3 to 5 issues. The full issue list, the fix roadmap, applied pull requests, and continuous monitoring are paid.
Checkout runs on Stripe and charges today in USD. Paying skips the waitlist. An invite code gets you into the beta, but it never grants Pro forever. Plan packaging may change during beta.
Accounts are isolated with row-level security, connector credentials are encrypted at rest, and GitHub access uses short-lived installation tokens with three narrow scopes.
For developers
Approved PRs on a crawrix branch.
Connect GitHub when you want the agent to open pull requests. Skip it and crawrix stays a diagnosis and a guided card in the dashboard.
Pull requestcrawrix/schema-localbusiness → main
fix(seo): add LocalBusiness JSON-LD to the homepage
Awaiting your approval+const localBusiness = {
+ "@context": "https://schema.org",
+ "@type": "LocalBusiness",
+ name: "Northline Plumbing",
+ areaServed: "Manchester",
+ sameAs: [profiles.google, profiles.facebook],
+}
- lintpassing
- typecheckpassing
- buildpassing
Pull requests, never main
Fixes arrive on a crawrix/... branch with Conventional Commits titles like fix(seo): .... Merging marks the linked findings resolved.
Checked before you see it
A sandbox can lint, typecheck, and build the branch first. The server stamps the pass or fail on the PR.
Narrow GitHub scopes
The GitHub App asks for Contents read and write, Metadata read, Pull requests write. Tokens are short-lived installation tokens.
CMS writes stay metadata-class
Where a CMS is connected: metadata fields, robots, llms.txt, schema, redirects. The previous value is logged and one click reverts it. Page body HTML is not rewritten.
Frameworks
- Next.js
- Astro
- SvelteKit
- Nuxt
- Remix
- React
- Vue
- Angular
- Gatsby
- Hugo
- Jekyll
- Eleventy
- Docusaurus
crawrix detects all of these. Full knowledge packs exist today for Next.js, where it writes confident PRs, and Astro, where it recommends first. Everything else falls back to generic SEO advice. Not every stack gets an auto-fix, and we would rather say so here than surprise you later.
Where it runs
Web chat today, more on the way.
The same scoring core is meant to run everywhere. Only the web chat is live, so that is the only one we are selling.
- Web chatThis is the product today.Live
- CLInpm i -g crawrix. Hosted mode on Max and above.Coming soon
- Hosted MCPFor Cursor and Claude Code, on Max and above.Coming soon
- WordPress pluginA thin client of the hosted service. Keeps your existing SEO plugin.Coming soon
- Shopify, Wix, WooCommerce, WebflowIn progress. There are no marketplace listings yet, on any platform.Coming soon
Guardrails
What is true, and what we will not claim.
crawrix is in private beta and the service is provided as is. Here is what it does today, and what it will never pretend to do.
True today
- Every write needs your approval. crawrix never pushes to main.
- You start with a URL. Connect GitHub only when you want PRs.
- Scores are best-effort diagnostics, and the formulas behind them are published.
- Multi-tenant isolation with row-level security. Connector and plugin credentials are encrypted at rest.
- The WordPress plugin sends nothing until an admin clicks Connect. No visitor analytics, comments, or orders.
- The service is provided as is during the beta period.
What we do not do
- We do not write blog posts or buy links.
- We do not guarantee Google rankings or AI citations.
- We do not track your position inside Google AI Overviews or Copilot. We only check whether their crawlers are allowed.
- No Search Console or Bing performance charts.
- No free trial. There is a free scan instead, and it needs no card.
- Nothing ships without your approval, and not every stack gets an automatic PR.
- No agency SKU and no white-label. Max covers 10 sites.
- No case studies or customer logos. This is a private beta.
Questions
What owners ask before they start
Straight answers on approvals, GitHub, trials, and what crawrix refuses to promise.
01Do you change my site without asking?
Never. Every write needs your approval, and we never push to main.
02Do I need GitHub?
No. Start with a URL. Connect GitHub only when you want pull requests.
03How is this different from Ahrefs or an agency?
Tools list keywords. Agencies sell hours. We measure AI visibility, rank what to do, and optionally ship approved technical fixes.
04Do you write content or buy links?
No.
05Is there a free trial?
No. There is a free scan with no card, and paid plans are monthly in USD. Paying charges you today and skips the waitlist.
06Can you tell me where I rank inside Google AI Overviews?
No. We check whether those crawlers are allowed to read your site. Anything beyond that is not built, so we do not claim it.
07I run several sites. What covers that?
Pro covers one site and extra domains are $35 a month each. Max covers up to 10. Past that, write to us. There is no agency plan or white-label.
08How do I get access?
Join the private beta waitlist for an invite code, or start a paid plan and skip the queue. Plan packaging may change during beta.
Free scan · private beta · or start Pro now
Stop guessing what AI says about you.
Run the free scan to see where you stand. Paid plans fix it, one approval at a time.
Private beta. We email a one-time invite code when your spot opens. No card required for the scan.
Questions first? support@crawrix.ai