Automated bots spam open source READMEs with AI project links

Bots are turning pull requests into ad space

Open source maintainers are dealing with a new strain of spam: automated accounts that open pull requests for no reason other than to plant promotional links to AI projects inside README files. According to a post on dev.to, the pattern has been visible since at least mid-September, and one bot was observed filing roughly three pull requests a day against repositories hosted on GitHub.

Individually, each request looks unremarkable. Aggregated across the ecosystem, they add up to a steady stream of junk that maintainers have to open, read and close before getting back to real contributions.

How the tactic works

The dev.to write-up breaks the campaign into simple mechanics. A script or AI-assisted tool scans for open repositories, locates the README, and opens a pull request that inserts the promotional link. The README is the deliberate target because it functions as the front page of a project, so anything placed there inherits the project's established audience.

The volume is calibrated to slip under the radar. A few requests per day spread across different repositories is not enough to trip GitHub's rate limits or basic spam checks, according to the post, but it is enough to become a chronic nuisance when repeated across many projects.

The incentive is visibility

The author frames the motivation as straightforward self-promotion. In a crowded AI landscape, attention is the scarce resource, and an established open source project offers free reach to anyone willing to hijack it. The cost side is equally lopsided: automation tools are cheap and accessible, while the burden of screening the resulting spam falls entirely on unpaid maintainers.

The post also flags an escalation risk. As language models improve, bots could start writing pull request descriptions that read as genuine help, presenting an inserted link as a useful resource rather than an obvious plug. At that point the failure mode shifts from platforms not filtering at all to filters being unable to tell machine-generated contributions from human ones.

Countermeasures under discussion

The dev.to article compares three possible responses:

• Enhanced spam filters: rated as effective at catching repetitive submission patterns, but with the caveat that poorly tuned filters could block legitimate contributions.
• CAPTCHA challenges on pull requests: judged only moderately useful, since they add friction for human contributors and AI-assisted bots may be able to solve them.
• Community moderation tooling: considered effective because it puts flagging and blocking power in maintainers' hands, though it depends on people actually having the time to use it.

The recommended approach is a combination of better filtering and community moderation, with AI-driven detection models held in reserve for the scenario where bots become good enough at mimicking humans to defeat conventional filters. The underlying point is that no single fix is permanent; detection has to keep evolving alongside the automation it is trying to catch.

Why it matters

Maintainer attention is the scarcest resource in open source, and spam campaigns spend it without permission. Every pull request that turns out to be a promotion is time taken away from code review, security work and mentoring new contributors. The dev.to post warns that if this behavior is left unchecked, it does more than annoy people: repeated spamming erodes trust in the contribution process itself, which is the foundation these projects are built on.

The timing makes the problem sharper. The same wave of accessible AI tooling that is lowering the barrier for legitimate automation is also lowering it for this kind of abuse, and the economics favor the spammer. Unless platforms introduce stronger defaults or the community builds shared defenses, each maintainer is left solving the same problem alone, and the busiest projects, the ones spam targets find most attractive, have the least spare capacity to fight it.