AI training decision is undecided

Why it matters

Keep training intentions separate from search access and from the permissions robots.txt currently declares.

What to do

1. Review the recorded decision and original robots lines in the finding.
2. In Settings → Detection, record whether you intend to allow or restrict training. Saving this intention does not edit robots.txt.
3. To change permissions, edit the relevant robots groups separately and run another complete audit.

When to leave it as it is: If unrestricted permissions are intentional, record that intention. Dismissing a finding does not record a training decision.

What we read

The settings frozen for this audit, one complete direct robots.txt from the registered origin and the admitted reviewed bot-purpose reference. Later settings do not reinterpret this audit.

The reviewed training crawlers currently include Amazonbot, Applebot, ClaudeBot and GPTBot. Applebot-Extended and Google-Extended are assessed separately as usage controls, not counted as crawlers. Public-dataset or unverified tokens are outside the cohort.

Only explicit undecided with completely known unrestricted permissions and no resolved substantive named training configuration gives this notice. Missing intent, raw or empty named groups, 404 robots, partial wildcard restrictions or unresolved companions remain unknown.

Example & evidence

An explicit decision with unchanged robots permissions; not evidence of model training.

Recorded decision: Not decided yet. Robots.txt:

User-agent: *
Allow: /

After review: Intend to allow training. Robots.txt remains unchanged. The complete audit records a decision, not an access change.

Confirm & check again

Check for yourself: Compare the decision saved for the audit and all exact named and wildcard groups. A narrow /wp-admin/ restriction does not prove that public-content training access was blocked.

Check again: Full audit — Run a new complete audit of the same site after recording a decision or making a resolved explicit robots configuration. A fully checked recorded decision can end this notice without proving an access change. Partial or unavailable facts cannot prove a fix; deleting the decision leaves it unknown.

Limits & sources

This reference is a reviewed subset, not an inventory of all bots. Permissions do not establish actual requests, model training or publisher compliance.

The setting records an intention. It publishes no opt-out, changes no website, infers nothing from dismissal and does not permanently silence findings.

Older audits without the field remain unrecorded; current settings are never substituted into their evidence.

Confidence: High The explicit original-run decision is bound to captured permissions and reviewed metadata. Actual model use is not asserted.

Sources

• Amazon, Amazon crawler documentation — Documents reviewed crawler purposes or usage controls and robots applicability, not actual fetching or training. Checked 9 September 2026.
• Google, Google common crawlers and usage controls — Documents reviewed crawler purposes or usage controls and robots applicability, not actual fetching or training. Checked 9 September 2026.
• Google, Google robots.txt documentation — Documents reviewed crawler purposes or usage controls and robots applicability, not actual fetching or training. Checked 9 September 2026.
• OpenAI, OpenAI bots — Documents reviewed crawler purposes or usage controls and robots applicability, not actual fetching or training. Checked 9 September 2026.
• Apple, Applebot and Applebot-Extended — Documents reviewed crawler purposes or usage controls and robots applicability, not actual fetching or training. Checked 9 September 2026.
• Anthropic, Anthropic crawler controls — Documents reviewed crawler purposes or usage controls and robots applicability, not actual fetching or training. Checked 9 September 2026.

Editorial review: 2 November 2026