Brief
AI agents are outpacing humans in reading documentation, raising accuracy concerns
Fortune notes that AI agents accessed documentation sites far more often than people in August, highlighting risks from stale or contradictory content.
By Felo News Desk · Published
In August, AI agents generated 257 million requests to documentation sites hosted by Mintlify, more than the 131 million page loads recorded from human users, Fortune reported. The surge means agents now read company knowledge at nearly twice the speed of people, yet many firms still write docs for a human audience.
What happened
Mintlify’s data shows AI agents queried its documentation platform 257 million times in August, while human developers accessed the same sites 131 million times. The agents use the retrieved information to answer queries, generate code, or provide support, often without the human habit of double‑checking the source.
What the reports add
Fortune explains that developers are increasingly delegating tasks to agents—such as wiring APIs, comparing vendors, or troubleshooting payments—so the agents rely heavily on existing docs. The article warns that outdated guides or pricing pages that conflict with current help‑center content can be copied verbatim by agents, propagating errors at scale.
What was said
Fortune quotes its own analysis, noting that 66 % of developers cite “AI output that is close to correct but still wrong” as their top frustration, attributing much of the problem to the quality of the underlying documentation.
How it came about
The shift stems from the rise of AI‑driven agents that treat documentation as a data source rather than a narrative for humans. Mintlify, the company behind the platform, highlights the need for new infrastructure to designate authoritative sources, synchronize docs with product changes, and monitor agent search failures as signals of misinformation.
Key facts
- AI agents made 257 million documentation requests in August, compared with 131 million from humans. (fortune.com)
- 66 % of developers say near‑correct but wrong AI output is their biggest frustration. (fortune.com)
- Outdated or contradictory docs can be copied by agents thousands of times before errors are caught. (fortune.com)
Sources
- [1] fortune.com — originally reported as “Nobody has been in charge for decades of the single most important part of the AI revolution: documentation”
