Agent AcceptedState of AI agent access
How 260 companies treat consumer AI agents, read from their robots.txt files and policy pages. The numbers update as we re-check; the latest check was on .
- Companies tracked
- 260
- With a clear policy
- 79
- Of those, block AI agents
- 59%
- Explicitly allow them
- 10
What the data says
- Of the 79 companies with a clear policy, 47 (59%) block AI agents, 10 allow them and 22 set conditions or limits.
- News is the most closed sector: 22 of 27 companies with a finding (81%) block AI agents.
- Perplexity (Perplexity) is the AI product blocked most often: 44 companies block it, out of 55 with a signal for it.
- 181 of 260 companies have no finding yet. "No policy found" does not mean a company allows agents.
By industry
Companies with a finding in each industry, split by verdict. Open an industry to see its companies.
- Blocked
- Conditional
- Selective
- Conflicting
- Allowed
| Industry | Verdicts | Block | With a finding | Tracked |
|---|---|---|---|---|
| News | Blocked 22, Selective 3, Allowed 2 | 22 (81%) | 27 | 43 |
| Retail | Blocked 7, Conditional 9, Allowed 4 | 7 (35%) | 20 | 94 |
| Social | Blocked 7, Selective 2 | 7 (78%) | 9 | 15 |
| Streaming | Blocked 3, Conditional 1, Selective 1 | 3 (60%) | 5 | 16 |
| Search | Blocked 2, Conditional 1, Allowed 1 | 2 (50%) | 4 | 11 |
| Travel | Blocked 1, Conditional 2, Allowed 1 | 1 (25%) | 4 | 13 |
| Banking | Blocked 3 | 3 (100%) | 3 | 12 |
| Jobs | Conditional 1, Allowed 1 | 0 (0%) | 2 | 3 |
| Food | Selective 1 | 0 (0%) | 1 | 5 |
| Investing | Blocked 1 | 1 (100%) | 1 | 3 |
| Local | Blocked 1 | 1 (100%) | 1 | 1 |
| Real Estate | Conditional 1 | 0 (0%) | 1 | 3 |
| Software | Allowed 1 | 0 (0%) | 1 | 4 |
No finding yet: Telecom (19), Gaming (7), Insurance (3), Payments (3), Ticketing (3), Shipping (2).
By AI agent
How many companies block, limit or allow each AI product. Only companies with a signal for the product are counted.
| AI product | Verdicts | Block | Limit | Allow | With a signal |
|---|---|---|---|---|---|
| Perplexity | Blocked 44, Conditional 7, Allowed 4 | 44 | 7 | 4 | 55 |
| Claude | Blocked 42, Conditional 13, Allowed 4 | 42 | 13 | 4 | 59 |
| Meta AI | Blocked 40, Conditional 5, Allowed 1 | 40 | 5 | 1 | 46 |
| ChatGPT | Blocked 36, Conditional 21, Allowed 9 | 36 | 21 | 9 | 66 |
| Gemini | Blocked 24, Conditional 4, Allowed 1 | 24 | 4 | 1 | 29 |
| Meta Muse | Blocked 8, Conditional 2, Allowed 1 | 8 | 2 | 1 | 11 |
How to read this
- A verdict comes from the company's own words (its terms or an agent policy page) or from the AI agent rules in its robots.txt. Every company page shows the evidence and the date.
- Silence is not permission. Most sites say nothing about AI agents, and a site that welcomes them usually has no rule at all. So blocking is far easier to detect than allowing, and "Allowed" will always look rare.
- "Limit" counts conditional verdicts. A rule in robots.txt is a request; some vendors say their agents may ignore it.
- 11 agent-rule changes were detected in robots.txt files in the last 30 days. See the news log.
- Coverage is still growing: the full method is on the methodology page.
Cite as "Agent Accepted, https://agentaccepted.com/insights". The data is licensed under CC BY-NC 4.0; see the terms. For bulk access or commercial use, contact us.