Intercom
Intercom/Support
- articles
- 5,384
- over six months old
- 37.5%
- languages
- 6
As of Aug 2, 2026
Read the reportAs of Aug 2, 2026
This report comes from a public crawl of the Zendesk help center. It covers 38,231 articles across English, Italian, Chinese, Japanese, French, and Portuguese. Every number is read from the sitemap and the pages themselves, none of it is estimated.

Health score
fair
out of 100
Platform
Zendesk Guide
Articles total
38,231
Categories
278
Languages
9
Primary language: English
Domain
support.zendesk.com
Own domain
Sitemap
40,675 entries
Sitemap index
One tile per article, sorted from freshest to most outdated.
6.1%6.1 percent of the English articles are older than six months
Half the library is under 54 days old. The other half stretches back 1,292 days.
One tile stands for 4 articles, so the map for all 4,587 of them fits on screen.
Articles changed per month over the last 24 months. Each bar counts the articles whose last change falls in that month.
4,569 articles changed over the last 24 months. The busiest month was July 2026, with 1,705 articles.
A weighted value across five areas. Anything we could not measure is left out rather than counted as a zero.
Every statement names the number it came from.
1 finding
There is no search field on the help center's front page.
7 findings
57 articles have not been touched in over a year, the oldest for 1,292 days.
On May 1, 2026, 821 articles share the same change date, which is 17.9 percent of the library. That points at a migration rather than at real edits.
100 percent of the article URLs carry a number instead of a readable title. That is the platform's doing.
79 percent of the images have no alt text, which makes them invisible to screen readers and to image search.
The articles offer no way to give feedback, so there is no signal about which article is not working.
There is no llms.txt, so AI systems get no curated entry point into the documentation.
Only 0 percent of the sampled pages show a change date, so readers cannot tell whether a guide still applies.
2 findings
Only 6.1 percent of the articles are older than six months. This library is being maintained.
All 9 AI crawlers we checked are allowed to read this help center.
Coverage relative to the primary language. Lag measures how many days a translation trails its source.
| Language | Articles | Coverage | Lag |
|---|---|---|---|
| EnglishPrimary | 4,587 | 100% | no data |
| Italian | 4,305 | 93.9% | 0 days |
| Chinese | 4,305 | 93.9% | 0 days |
| Japanese | 4,254 | 92.7% | 0 days |
| French | 4,242 | 92.5% | 0 days |
| Portuguese | 4,237 | 92.4% | 0 days |
| German | 4,215 | 91.9% | 0 days |
| Spanish | 4,215 | 91.9% | 0 days |
| Korean | 3,871 | 84.4% | 0 days |
Whether ChatGPT, Claude and Perplexity are allowed to read and cite this help center at all, and whether they can tell how current an article is.
Crawler access
llms.txt
absent
Structured data
Organization, SoftwareApplication, TechArticle, WebPage, WebSite
Visible change date
on 0 percent of the sampled pages
Based on a sample of 40 article pages. The values describe that sample, not every single page.
Up to 30 articles, sorted by when they last changed.
Drop in the URL, we crawl it and send the full report by email. Usually within two working days. Whether it gets published is your call.
These teams document with HappySupport
We read the help center's public sitemap, assign every URL to a language and evaluate the change date it declares. We then fetch a sample of article pages spread across the whole library and read titles, descriptions, structured data and interface controls off them. All of it runs under a clearly identifiable user agent, with pauses between requests.
Limits: change dates come from the underlying system and may have been reset in bulk during a migration. We call those days out separately. Pages that block us are left out of the sample. Nothing in this report says anything about the editorial quality of the articles.
This report can be wrong
Everything here is measured automatically from public pages, so a report can misread a help center or miss part of it. If a number does not match what you know about your own help center, write to contact@happysupport.ai and we will check it and correct the report.
Zendesk is a trademark of its respective owner. This report was produced independently and is not affiliated with Zendesk.
Other help centers we have measured.
Intercom/Support
As of Aug 2, 2026
Read the reportZendesk Guide/Collaboration
As of Aug 2, 2026
Read the reportIntercom/Other
As of Aug 2, 2026
Read the reportStart for free and publish your first article today.
