Evidence-first AI discovery audit

Check whether AI systems can access and interpret your public website

OpenForBots reviews crawler policies, public retrieval, initial HTML, business clarity, and supporting evidence. See what was found, what it may mean, and what to improve—without speculative visibility scores.

  • Evidence-labelled findings
  • Search and training controls separated
  • Practical recommended actions
  • No ranking guarantees
Preview your audit Enter a public domain and choose the policy outcome you want to review.

Current status: the preview does not contact the submitted website. No account or email is required.

See the planned report format ↓

Crawler access summary

Example audit result

Example data
OAI-SearchBot Search discovery
Open
GPTBot Model training
Blocked
Claude-SearchBot Search discovery
Open
Google-Extended Data-use control
Review

Statuses reflect crawler purpose. Blocking training is not automatically a search-access problem.

Planned Release 1 flow

From public domain to practical action

The audit is designed to keep directly observed facts separate from interpretation and external outcomes.

  1. 01

    Submit a public domain

    Choose whether you want balanced search access, maximum recognised access, or an observation-only review.

  2. 02

    Inspect policy and public content

    Check crawler directives, public responses, initial HTML, business signals, and evidence readiness within defined technical limits.

  3. 03

    Review evidence and actions

    Receive the finding, source or observation, plain-language meaning, limitation, recommended fix, and verification condition.

Illustrative report format

See how one finding will be explained

This synthetic fixture demonstrates the report anatomy. It is not a live scan or a claim about a real website.

Crawler policy

OAI-SearchBot is not explicitly blocked in the example policy

Observed
Evidence
The version-controlled fixture contains no matching disallow rule for OAI-SearchBot.
Meaning
The tested robots policy does not appear to restrict the documented OpenAI search crawler.
Does not prove
That OpenAI indexed, retrieved, cited, ranked, or recommended the website.
Recommended action
No policy change is required unless the site owner intended a different access rule.
Verification condition
Recheck after any robots.txt, CDN, firewall, or hosting-policy change.

What gets evaluated

A practical audit, not a generic AI score

Each category produces inspectable evidence and a next action rather than an unexplained visibility percentage.

01

Crawler policy

Checks recognised tokens separately by search, training, retrieval, and data-use purpose.

Output: matched directives, purpose interpretation, conflicts, and policy action.

02

Public retrieval

Examines public responses and initial HTML within deterministic request limits.

Output: response evidence, retrieval barriers, and bounded technical fixes.

03

Business clarity

Checks whether the company, category, audience, offer, and supporting facts are plainly exposed.

Output: extracted signals, missing context, and copy or structure improvements.

04

Evidence quality

Separates observations and provider documentation from simulation, heuristics, inference, and manual review.

Output: evidence label, explicit limitation, and verification path.

Crawler purpose matters

Search access and model training are separate decisions

Blocking a training crawler may be intentional. OpenForBots does not automatically turn that choice into a search-visibility failure.

Read the crawler distinction

Search discovery

Controls that may affect retrieval for search or answer experiences.

Model training

Controls related to using public content to develop future models.

User-triggered retrieval

Requests initiated by a user action, potentially governed by different rules.

Data-use controls

Provider controls that govern use rather than acting as standalone page crawlers.

Trust through explicit limits

Every finding explains how it was established

Evidence type is shown separately from downstream impact. No access result is presented as a guarantee of indexing, citation, ranking, recommendation, or traffic.

Observed
Observed

Directly retrieved or parsed during a deterministic check.

Documented
Documented

Supported by a current provider-maintained or standards source.

Simulated
Simulated

Produced by a controlled request that does not prove production crawler behaviour.

Heuristic
Heuristic

A bounded rule of thumb with visible limitations.

Inferred
Inferred

A reasoned conclusion rather than a directly observed fact.

Manual check
Manual check

Something the automated audit cannot determine responsibly.

Who it helps

Built for people making real website decisions

Different users need different policy outcomes, but all need the evidence and limitation behind a recommendation.

Founders and small businesses

Understand whether the public site explains the product clearly and whether obvious crawler barriers exist.

Developers and technical teams

Review robots directives, retrieval behavior, initial HTML, and implementation-oriented fixes.

SEO professionals and agencies

Create evidence-backed recommendations without presenting speculative AI visibility scores as facts.

Publishers and content teams

Separate search access, user retrieval, model training, and data-use choices.

Provider-sourced reference data

Latest verified crawler references

Each entry records the crawler token, purpose, policy notes, misconceptions, official source, and verification date.

Common questions

What the audit can—and cannot—tell you

Is the OpenForBots scanner live?

Not yet. The current audit flow previews the inputs, policy choices, evidence model, and planned report format without contacting the submitted website. The secure Release 1 scanner is being developed separately.

Does blocking GPTBot block ChatGPT Search?

Not by itself. OpenAI documents GPTBot and OAI-SearchBot as separate controls with different purposes. OpenForBots evaluates recognised crawler tokens individually instead of treating every OpenAI-related rule as the same decision.

Is llms.txt required for AI search visibility?

No universal requirement has been established. OpenForBots may inspect the file when it exists, but its absence is not treated as a critical failure or a guaranteed visibility problem.

Can OpenForBots guarantee AI citations or rankings?

No. OpenForBots can evaluate observable readiness conditions such as crawler policy, public retrieval, initial HTML, business clarity, and evidence signals. It cannot see a provider’s complete private index, ranking logic, retrieval process, or recommendation system.

Does the Release 1 audit execute JavaScript?

The planned Release 1 deterministic scan focuses on initial public responses and does not run an unrestricted headless browser. A heavy client-rendering dependency may be reported as a heuristic, not as proof that a particular provider cannot render the page.

What does business clarity mean?

It means checking whether the public page clearly exposes signals such as the organisation, product category, intended audience, primary offer, title, heading, opening explanation, and supporting evidence. It does not claim to observe what a private model internally understands.

Start with the observable layers

Preview the audit for your website

Review the input flow and policy model now. The page clearly labels which functionality is still under development.