Find out what ChatGPT and Claude are allowed to read
The GEO Intelligence API audits a page's AI-search readiness as clean JSON.
6 active endpoints. Every call is 1 credit.
- POST/geo-intel/v1/check
- POST/geo-intel/v1/crawlers
- POST/geo-intel/v1/llms_txt
- POST/geo-intel/v1/schema
- POST/geo-intel/v1/score
- POST/geo-intel/v1/batch
What AI Search Visibility endpoints does ReefAPI ship?
6 live read endpoints. Read-only data API: no writes, no account actions, no dashboard access on the target site.
AI Search Visibility API
3 of 6 endpoints, ready to run
One citability score with a grade, broken into crawler access, structured data, llms.txt, meta signals and crawlability — plus the specific recommendations to raise it.
{ "ok": true, "meta": { "api": "geo-intel", "endpoint": "score", "mode": "live", "latency_ms": 3280.7, "record_count": 1, "cache_hit": false }, "data": { "url": "https://reefapi.com", "final_url": "https://reefapi.com", "score": 90, "grade": "A", "breakdown": { "crawler_access": { "score": 40, "max": 40 }, "structured_data": { "score": 15.5, "max": 25 }, "llms_txt": { "score": 15, "max": 15 }, "meta_signals": { "score": 11.5, "max": 12 }, "crawlability": { "score": 8, "max": 8 } }, "recommendations": [], "signals": { "critical_bots_blocked": [], "citation_bots_blocked": [], "blanket_block_all": false, "llms_txt_present": true, "has_structured_data": true, "high_value_schema_types": [ "Organization" ], "noindex": false, "has_canonical": true, "has_opengraph": true, "sitemap_present": true, "likely_js_app": false } } }
How the AI Search Visibility API works
AI Search Visibility is a normal ReefAPI surface — the same four rules that hold for every other engine on the key.
No OAuth app, no request signing, no per-site account. One key covers all 184 engines.
Every route is a POST with a JSON body. Parameters are validated against the published schema before anything is charged.
Credits, not seats. Failed and blocked calls are never charged, and cache hits cost nothing.
One envelope everywhere. meta carries latency_ms, record_count and the endpoint that answered.
Finding out why an assistant will not cite you
Ranking in a search engine and being usable by an assistant are different problems. A site can be perfectly indexed and still be unreadable to the crawlers that feed AI answers.
{"url": "https://example.com"}Tests dozens of AI crawler tokens against your live robots.txt and reports which are blocked, separating the ones that cite sources from the ones that only train.
{"url": "https://example.com"}Then the whole picture as a graded score, with the recommendations that would move it.
The distinction the report makes is the useful one: blocking a training crawler is a choice, and blocking a citation crawler is lost attribution. Most sites do not know which they have done.
curl -X POST https://api.reefapi.com/geo-intel/v1/score \
-H "x-api-key: $REEF_KEY" \
-H "content-type: application/json" \
-d '{"url":"https://github.com"}'{
"ok": true,
"data": { … },
"meta": {
"api": "geo-intel",
"endpoint": "score",
"mode": "live",
"latency_ms": …,
"record_count": …
},
"error": null
}The five scoring pillars and what each one reads
GEO here means Generative Engine Optimization, not geolocation. The only input is a URL and the output is a 0-100 AI-readiness score built from five pillars whose maximums add to 100. The breakdown block returns each pillar's earned score against its max, and recommendations name the pillar, the priority and the exact points_recoverable, so you can reconstruct the arithmetic yourself.
| Pillar | Max points | What it reads, with a measured example |
|---|---|---|
| crawler_access | 40 | robots.txt verdict for 73 AI bot tokens using Google longest-match; www.nytimes.com blocked 29 of them, 7 of which are flagged critical |
| structured_data | 25 | JSON-LD blocks and @graph types; github.com scored 6.0 because it ships RDFa but no JSON-LD |
| llms_txt | 15 | /llms.txt plus /llms-full.txt with the content type sniffed so an HTML 404 cannot pass; docs.anthropic.com returned a valid 72,234-byte file with 689 links |
| meta_signals | 12 | canonical, OpenGraph and the robots meta tag; github.com scored 11.5 of 12 |
| crawlability | 8 | a declared sitemap.xml, the text ratio and likely_js_app; github.com lost half of it for having no sitemap |
Measured score-to-grade pairs from live runs: 90.0 = A (vercel.com, reefapi.com), 80.5 and 76.5 = B (docs.anthropic.com, github.com), 64.0 = C (developer.mozilla.org), 50.6, 49.0 and 45.0 = D (bbc.com, en.wikipedia.org, stripe.com), 5.0 = F (reddit.com). score is a float, not an integer.
What is measured, and what a score can honestly claim
Measured against live sites. The last row is the limit of what any tool of this kind can tell you.
Every known AI crawler token is evaluated against your live robots.txt for the path you name, and each result reports the rule that decided it. It is an evaluation of your file, not a lookup of what a crawler usually does — so a rule that accidentally matches is visible, with the matching line quoted back.
Each bot carries its operator, the product it feeds and whether that product cites its sources. Blocking a crawler that never attributes you costs nothing; blocking one that puts a link next to the answer costs the attribution. Collapsing both into one 'AI bots' number is the mistake this breakdown exists to prevent.
The check reports whether the file exists, whether the content type is really plain text, whether the body is secretly an HTML error page, and how it is structured — section count, link count and the specific findings. A 200 response that returns your styled 404 page is the common failure here, and status alone will not catch it.
The grade breaks into crawler access, structured data, llms.txt, meta signals and crawlability, each with its own points and ceiling. Two sites with the same total can need completely different work, and the breakdown is what says which.
Everything here is about whether an assistant can reach and parse your pages. Whether it then chooses to cite you depends on your content, your reputation and where else you are mentioned — none of which a crawler check can see. Treat a high score as the precondition being met, not as the outcome being achieved.
What people build with AI Search Visibility
The jobs this data is most often used for.
endpoints
credit per call
SEO/GEO tools call check to grade a page's readiness for AI search.
Agencies use the recommendations to prioritize fixes for a client.
Monitors use batch to audit a whole site's pages.
What AI Search Visibility data costs
The cheapest call here is 1 credit, so $15/mo (Pro) buys 10,000 of them — $1.50 per 1,000 credits. Credits roll over and never expire, and failed or blocked calls are not charged.
Full pricing →- 1,000 free credits on signup, no card
- One key, all 184 APIs, one credit pool
- Failed and blocked calls are never charged
- Credits roll over and never expire
Call it in two lines
Sign up, get 1,000 credits and one key that works on every engine. Then this is the whole protocol.
curl -X POST https://api.reefapi.com/geo-intel/v1/score \
-H "x-api-key: $REEF_KEY" \
-H "content-type: application/json" \
-d '{"url":"https://github.com"}'import requests
r = requests.post(
"https://api.reefapi.com/geo-intel/v1/score",
headers={"x-api-key": REEF_KEY},
json={
"url": "https://github.com"
},
)
print(r.json()["data"])Have a question? We got answers.
The questions people actually ask before wiring up AI Search Visibility.
Get a free key →Is this a geolocation API?▾
No. geo-intel never accepts a coordinate, an IP or a postal code, and its only required parameter is a URL. For coordinates and postal codes use geo-postal, and for locating an IP address use ip-geolocation. The name is short for Generative Engine Optimization, which is search-visibility work aimed at ChatGPT, Claude, Perplexity and Google AI Overviews rather than at a map.
Which bots count as critical, and why does blocking GPTBot not hurt my score?▾
Every bot row carries cites and critical booleans alongside category, which is one of search, agent, assistant, training or scraper. On www.nytimes.com the 7 critical_blocked entries were OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User and Google-Extended, all bots that can produce a citation. GPTBot was blocked on the same site but is category training with cites false, so it does not appear in critical_blocked. The score punishes blocking bots that could cite you, not the ones that only train.
The response says has_structured_data true but the schema pillar still lost points. Why?▾
has_structured_data covers microdata and RDFa as well as JSON-LD, while the pillar rewards JSON-LD specifically. github.com returned has_structured_data true and scored 6.0 of 25, with the recommendation "only microdata/RDFa found - JSON-LD is the AI-preferred format; add it" and points_recoverable 19.0. Read citable_types and high_value_types instead of the boolean: on nytimes.com those were NewsMediaOrganization and WebSite out of four detected types, with type_counts showing 43 ListItem entries.
Does it follow redirects, and can I point it at an internal host?▾
follow_redirects defaults to true and each hop is validated independently, and the response returns both url and final_url so you can see where the analysis actually landed. Private, internal and cloud-metadata targets are refused: a call for http://169.254.169.254/latest/meta-data/ returned INVALID_PARAM with the message "SSRF-blocked: private/internal IP" and a detail block naming the host and its resolved addresses.
What does batch give up compared with calling check per URL?▾
batch takes up to 20 URLs and scored three of them concurrently in a single 3.4 second call, but each result carries only url, final_url, score, grade, critical_blocked, llms_txt_present and has_structured_data. There is no per-bot table, no llms.txt parse and no recommendations. Use batch to rank a portfolio, then call crawlers or check on the ones that scored badly.
What is robots_status, and what if robots.txt is unreadable?▾
robots_status is the literal HTTP status of the robots.txt fetch; www.nytimes.com answered 206 rather than 200 and the parse still succeeded. Alongside it you get robots_present, parse_errors as an array, user_agents_in_robots listing the agent blocks that actually appear in the file, and blanket_block_all for a global Disallow of the site root. If parse_errors is non-empty, treat the per-bot allowed values as unreliable rather than as permission.
Can I test a specific page instead of the homepage?▾
Yes, with path, which defaults to "/". The check, crawlers and score actions evaluate robots.txt allow and deny rules against that exact path, so passing /blog/some-post tells you whether AI bots may fetch that article rather than whether they may fetch the site root. Sites frequently allow the root and disallow a content directory, and only the path form catches that.
What does llms_txt check that a plain fetch would not?▾
It sniffs the content type so an HTML 404 page served at /llms.txt cannot be counted as present. A measured docs.anthropic.com response returned present true, valid true, status 200, content_type text/plain, looks_like_html false, bytes 72234, title "Anthropic Developer Documentation", section_count 3, link_count 689 and an empty findings array. llms_full_txt is reported separately with its own status and URL.
What is the AI Search Visibility API?▾
AI Search Visibility API is a ReefAPI endpoint group for ai search visibility It returns live JSON through POST requests under /geo-intel/v1.
Is the AI Search Visibility API free to try?▾
Yes. ReefAPI starts with 1,000 free credits, no card required. AI Search Visibility calls use the same shared credit balance as every other ReefAPI engine.
Do I need an AI Search Visibility login or account?▾
No login to AI Search Visibility is needed for the API response. You call ReefAPI with your x-api-key header, and the playground can run live examples before you create a production key.
How fresh is the AI Search Visibility data?▾
The page example is captured from a live check call, and production requests fetch live data through ReefAPI rather than a static sample.
How many credits does the AI Search Visibility API use?▾
AI Search Visibility actions currently cost 1 credit per successful call. Failed or blocked calls are free, and all APIs draw from one credit pool.
Can I call AI Search Visibility from an AI assistant or MCP client?▾
Yes. Connect ReefAPI once through MCP and your assistant can call geo-intel actions with the same key, credit pool and JSON envelope used by normal REST requests.
19 Utilities & AI APIs on the same key
One key, one credit pool, one response envelope. If you are pulling AI Search Visibility, you are one call away from the rest of the category — no second contract, no second integration.
Need something this API does not do?
Name the endpoint, the field, or a source we do not carry yet. We ship new APIs every week and you would be first to get the key. Real people read every message and reply the same day.
Try it on your own data before you pay anything
The call above is the real endpoint, not a recording. A free key gives you 1,000 credits, the other 183 APIs, and the same envelope everywhere.
Endpoints, parameters and credit costs on this page are read from the live catalog and cannot drift from what the API accepts. Field notes were captured on 2026-08-30.