Get now

Page token budget

A page that does not fit an agent's context window gets truncated, so some of its content may be missed.

Standard
Recommended
100%Top 100 adoption97 of 97 sites implement it5/9OverviewGaining tractionReal support exists and adoption is growing.

Who supports it

  • ClaudeSupported (announced)

    Undated · last checked 9 Oct 2026: Claude API docs: Web fetch toolmax_content_tokens: 'If the fetched content exceeds this limit, the tool truncates it.' Docs give ~25,000 tokens for a 100 kB page. Browser use tool also caps read_page output at 50,000 characters.

  • GeminiPartial support

    Published 23 Sept 2026 · last checked 9 Oct 2026: Gemini API docs: URL contextGemini API URL context: 'The maximum size for content retrieved from a single URL is 34MB'; up to 20 URLs per request; retrieved content counts toward input tokens. API tool, not necessarily the consumer Gemini app.

  • Claude CodeSupported (announced)

    Published 5 Oct 2026 · last checked 9 Oct 2026: Claude Code CHANGELOG.md (anthropics/claude-code)v2.1.290 (npm release 2026-10-05): 'Fixed WebFetch silently dropping page text past 100,000 characters; it now says how much was unread and takes an offset to read on.' Matches this check's 100,000-character threshold.

24 others: no evidence yet. See every agent and browser

What it checks

The extension counts the characters of text on the page. It uses the rendered page when it can read it, and falls back to the text in the server’s HTML when it cannot. The check passes at 100,000 characters or fewer, which is roughly 25,000 tokens at four characters per token.

Past that point, an agent working within a context limit is likely to truncate the page, and whatever falls after the cut-off never gets read.

Results

Status When
Pass The page has 100,000 characters of text or fewer
Warn The page has more than 100,000 characters of text
Unmeasured The rendered page could not be read and the HTML could not be fetched

The result shows the character count and an approximate token count either way.

How to fix

Split oversized pages into focused documents, one topic each, and link them from an index page. Long reference pages, full changelogs and single-page documentation are the usual cases.

If you publish an llms.txt, list the smaller pages there so an agent can fetch only the one it needs.

Support by agent and browser

Announced means the vendor’s own docs or announcement say so. Published is the date on the vendor’s own page, where it has one. Checked is the last time our weekly research found that page still saying so.

AgentStatusSource
AI assistants
Claude AnthropicSupported (announced)

Claude API docs: Web fetch tool Undated · Checked max_content_tokens: 'If the fetched content exceeds this limit, the tool truncates it.' Docs give ~25,000 tokens for a 100 kB page. Browser use tool also caps read_page output at 50,000 characters.

Gemini GooglePartial support

partial Gemini API docs: URL context Published · Checked Gemini API URL context: 'The maximum size for content retrieved from a single URL is 34MB'; up to 20 URLs per request; retrieved content counts toward input tokens. API tool, not necessarily the consumer Gemini app.

Coding agents
Claude Code AnthropicSupported (announced)

Claude Code CHANGELOG.md (anthropics/claude-code) Published · Checked v2.1.290 (npm release 2026-10-05): 'Fixed WebFetch silently dropping page text past 100,000 characters; it now says how much was unread and takes an offset to read on.' Matches this check's 100,000-character threshold.

No evidence yet24
  • ChatGPT
  • Perplexity
  • Copilot
  • Meta AI
  • ChatGPT Atlas
  • Comet
  • Dia
  • Gemini in Chrome
  • Edge Copilot Mode
  • Cursor
  • Codex
  • GitHub Copilot
  • Windsurf
  • CCBot
  • Applebot
  • Amazonbot
  • Bytespider
  • Cohere
  • Mistral
  • DuckAssist
  • Chrome
  • Edge
  • Firefox
  • Safari

Names and marks belong to their owners and are used only to identify products. None of these vendors is affiliated with Sitebulb. Spotted something out of date? Send a correction.

Adoption on the top 100 sites

97 of 97 measured sites implement it

Scanned 23 Oct 2026 with the Sitebulb Labs extension, against the Tranco list (tranco-56Z8N). 3 sites couldn’t be measured (bot challenge or timeout) and are left out of the share.

9 Oct 2026: 100%12 Oct 2026: 100%23 Oct 2026: 100%
9 Oct 2026 100% → 23 Oct 2026 100%
Sites implementing it97
  1. #1 google.com
  2. #2 cloudflare.com
  3. #3 aws.amazon.com (via amazonaws.com)
  4. #4 facebook.com
  5. #5 gstatic.com
  6. #8 googleapis.com
  7. #9 microsoft.com
  8. #11 youtube.com
  9. #12 apple.com
  10. #13 instagram.com
  11. #15 mail.ru
  12. #16 x.com (via twitter.com)
  13. #18 linkedin.com
  14. #20 dzen.ru
  15. #23 googletagmanager.com
  16. #24 azure.microsoft.com (via azure.com)
  17. #26 m365.cloud.microsoft (via office.com)
  18. #27 outlook.live.com (via live.com)
  19. #32 github.com
  20. #33 wikipedia.org
  21. #37 fastly.com (via fastly.net)
  22. #39 whatsapp.com (via whatsapp.net)
  23. #42 bing.com
  24. #43 netflix.com
  25. #44 triplinkintl.com
  26. #45 googleusercontent.com
  27. #48 gandi.net
  28. #50 marketingplatform.google.com (via doubleclick.net)
  29. #52 uk.pinterest.com (via pinterest.com)
  30. #53 icloud.com
  31. #55 wordpress.org
  32. #56 teams.live.com (via skype.com)
  33. #57 cloudflare.net
  34. #59 goo.gl
  35. #60 digicert.com
  36. #61 adobe.com
  37. #63 tiktok.com
  38. #67 domains.google (via googledomains.com)
  39. #68 zoom.com (via zoom.us)
  40. #69 uk.yahoo.com (via yahoo.com)
  41. #70 sso.myfritz.net (via myfritz.net)
  42. #71 endpoints.news
  43. #75 roblox.com
  44. #76 opera.com
  45. #77 open.spotify.com (via spotify.com)
  46. #80 baidu.com
  47. #81 tiktokv.com
  48. #83 ya.ru (via yandex.net)
  49. #84 nginx.org
  50. #85 ntp.org
  51. #86 msn.com
  52. #87 qq.com
  53. #90 vimeo.com
  54. #91 primevideo.com (via amazonvideo.com)
  55. #93 bunny.net (via b-cdn.net)
  56. #94 blogger.com (via blogspot.com)
  57. #95 mozilla.org
  58. #96 f5.com
  59. #98 telegram.org (via t.me)
  60. #99 store.epicgames.com (via epicgames.com)
  61. #101 wordpress.com
  62. #103 reddit.com
  63. #106 sentry.io
  64. #107 letsencrypt.org (via lencr.org)
  65. #109 ui.com
  66. #110 european-union.europa.eu (via europa.eu)
  67. #112 chatgpt.com
  68. #114 st.okcdn.ru (via okcdn.ru)
  69. #116 bitly.com (via bit.ly)
  70. #117 ipify.org
  71. #119 nic.ru
  72. #120 samsung.com
  73. #121 office365.com
  74. #122 intuit.com
  75. #123 aliyuncs.com
  76. #127 openai.com
  77. #128 one.one.one.one (via cloudflare-dns.com)
  78. #130 registrar-servers.com
  79. #131 pages.github.com (via github.io)
  80. #132 dropbox.com
  81. #133 skybridge.click
  82. #135 discord.com (via discord.gg)
  83. #136 developer.adobe.com (via adobe.io)
  84. #139 apache.org
  85. #140 android.com
  86. #141 nih.gov
  87. #144 vk.ru
  88. #147 reg.ru
  89. #150 t-online.de
  90. #151 archive.org
  91. #152 forms.gle
  92. #153 webex.com
  93. #154 snapchat.com
  94. #156 godaddy.com
  95. #158 kaspersky.co.uk (via kaspersky.com)
  96. #159 firebase.google.com (via crashlytics.com)
  97. #160 tumblr.com

Overview

5/9

Gaining traction

Real support exists and adoption is growing.

How this rating is worked out5 of 9 points
FactorPointsWhy
Vendor support1 / 31 vendor supporting, 1 partly or planned
Top-100 adoption3 / 3100% of measured sites
Spec maturity1 / 3Single-vendor spec

8 or more points is Established, 5 to 7 is Gaining traction, 2 to 4 is Speculative and under 2 is Stalled. A feature is also Stalled when more vendors have said no than yes. Read the spec.

Community insights

No community insights for this feature yet.

Quotes are from public posts, checked against the original before they appear here. Suggest an insight, or ask for yours to be removed.