CrawlDex
Check

State of the agent web

Agent-Friendly Websites Report

A live report on which websites and task categories are easiest for AI agents to complete with visible evidence.

Publication

Human reviewed
8,401
site-task rows
2,190
websites
21%
need a person nearby
68
mean AES

Last verified Jul 8, 2026. Based on 9,590 observations.

Foreword

A usable web is becoming a measurable web.

The clearest pattern in this board is not that agents are universally good or bad at websites. It is that sites with stable task state, clear confirmation pages, and documented user-present checkpoints produce more reliable agent outcomes. The remaining gap is practical: 21% of measured site-task rows still need a person nearby for authentication, payment, confirmation, or support handoff.

Category league table

Agent reliability by task category

Permalink

Each category uses the same board rows as the public task and ranking pages.

CategoryMean AESRowsNeed a person nearbyTop blocker
Government And Civic8210729%public url verified
Jobs And Careers829858%login walls
Developer, SaaS, And API8110870%login walls
Education815956%public url verified
Creator, Social, And Media799883%login walls
Business Operations7810184%login walls
Healthcare Admin7810159%login walls
Future Agent Tasks769970%login walls
Legal And Admin765054%public url verified
Customer Support7560100%required human handoff
Travel715895%login walls
Real Estate And Moving687990%public url verified
Finance And Bills6634100%login walls
Insurance6533100%required human handoff
Subscriptions6310452%public url verified
Food And Local Delivery5963100%login walls
Commerce566062%public url verified

Blocker prevalence

The blockers that appear most often

Permalink

Rows can have more than one blocker, so counts are affected site-task rows rather than a mutually exclusive split.

public_url_verified

public url verified

6787 rows

The task repeatedly runs into public url verified, which changes the expected agent path or prevents a clear final state.

final_action_requires_human

final action requires human

4629 rows

The task repeatedly runs into final action requires human, which changes the expected agent path or prevents a clear final state.

login_required_for_final_action

login required for final action

3794 rows

The task repeatedly runs into login required for final action, which changes the expected agent path or prevents a clear final state.

payment_or_fee_context

payment or fee context

3212 rows

The task repeatedly runs into payment or fee context, which changes the expected agent path or prevents a clear final state.

login_required

login walls

1395 rows

The site requires account authentication before the task can be measured or completed.

site_access_denied

site access denied

991 rows

The task repeatedly runs into site access denied, which changes the expected agent path or prevents a clear final state.

http_403

http 403

972 rows

The task repeatedly runs into http 403, which changes the expected agent path or prevents a clear final state.

sparse_public_text

sparse public text

757 rows

The task repeatedly runs into sparse public text, which changes the expected agent path or prevents a clear final state.

hidden_modal

hidden or obstructing modals

462 rows

A modal, interstitial, or overlay obscures the task path or changes the expected controls.

human_handoff_required

required human handoff

204 rows

The task reaches a step where the account holder, support staff, or another human must act.

Handoff analysis

21% of measured rows need a person nearby

Permalink

The handoff estimate is recomputed from rows with user-present blockers, login walls, payment approval, confirmation gaps, or explicit human handoff status.

CategoryNeed a person nearbyRowsLast verified
Food And Local Delivery100%63Jul 8, 2026
Customer Support100%60Jul 8, 2026
Finance And Bills100%34Jul 8, 2026
Insurance100%33Jul 8, 2026
Travel95%58Jul 8, 2026
Real Estate And Moving90%79Jul 8, 2026
Business Operations84%101Jul 8, 2026
Creator, Social, And Media83%98Jul 8, 2026

Site-task tables

Friendliest and hardest current rows

Permalink

These tables link back to the measured site-task pages so each claim stays inspectable.

Most-checked tasks

Priority site-task evidence behind this report

These pages are selected from evidence quality, consumer intent, and Search Console signals when available.

Methodology

How to read this report

Permalink

The report is generated from the same CrawlDex board API as rankings, blockers, and dataset pages.

Freshness

Report tables use the latest public board update: Jul 8, 2026.

Confidence

Counts come from 9,590 observations across 8,401 site-task rows.

Corrections

Negative claims include a dispute path so site owners and users can challenge stale or incorrect measurements.

Agent-Friendly Websites Report | CrawlDex