There's a quiet shift in how local development decisions get reported, and anyone who has stood at a Planning Board podium in the last year has probably noticed it. The traditional local reporter — the one who sat through four hours of testimony, read the resolution, and called the applicant for comment — has largely been replaced. In their place: a growing class of media platforms that scrape agendas, ingest video, run it through a large language model, and publish "coverage" within hours of a meeting ending.
No analysis. No fact-checking. No phone calls. Just AI-generated copy engineered to do one thing: get clicked.
The mechanics aren't a secret. A platform pulls the public meeting video, auto-transcribes it, feeds the transcript to a model with a prompt that effectively says "summarize this and make it engaging," and pushes the output through a CMS with a programmatically generated headline. The headline is the product. Everything beneath it is filler designed to keep a reader on the page long enough to register ad impressions.
The model doesn't know who the applicant is. It doesn't know the zoning history of the lot. It doesn't know that a board member's clarifying question wasn't an objection, or that the public commenter who spoke for eleven minutes lives four towns over. It just produces text that sounds like reporting.
A board chair asking "have you considered parking?" becomes "Board Slams Developer Over Parking Concerns." A neighbor expressing general support with one reservation becomes "Residents Raise Alarm." An applicant explaining a setback variance becomes "Developer Seeks Exemption From Rules."
Each rewrite is technically derived from something that happened. None of them describe what actually happened.
And the headlines are the worst of it, because the headlines are what circulate. Most people never click. They read the headline, form an impression, and move on. The impression is the only thing the platform was ever selling.
This is the part the platforms don't have to care about, and applicants can't ignore. Once the wrong facts are out, they are the story. The fabricated headline gets aggregated. The aggregator gets quoted. The quote ends up in the next article, which cites the first one as a source. Within a week, an invented characterization of a meeting is being referenced as if it were the record.
Corrections — when they happen at all — don't catch up. The original post stays indexed. Six months later it's still the first result when someone searches the project name. Lenders read it. Neighbors read it. Future tenants read it. The applicant spends the next round of approvals answering for something that was never said.
A workflow that ingests a four-hour meeting, generates a headline through a probabilistic text model, and publishes it without a human reading the underlying transcript is — at minimum — not in the business of verifying anything. The publisher has chosen a process that cannot, by design, catch its own errors before they reach an audience.
Whether that crosses the legal threshold for defamation in any given case is a question for a court. The posture, though, is plainly negligent. The platforms know the model hallucinates. They publish anyway. They publish because it's cheap to publish, and the occasional false statement about a private developer or a small business is, to them, just a cost of doing business at scale.
It's not a cost the developer or the business gets to pass along. It sits on their reputation.
If a platform is going to publish coverage of a public meeting, the bar shouldn't be high. A human should read the article before it goes live. The applicant should be contacted for comment. Factual claims should be traceable to the record. The headline should describe what happened, not what would get the most clicks.
None of that is a radical standard. It's the standard local newspapers held themselves to for a century. The fact that an automated pipeline can't meet it is a description of the pipeline, not of the standard.
Until that changes, the responsibility falls to the rest of us — applicants, neighbors, board members, anyone who was actually in the room — to point out, plainly and on the record, when the coverage doesn't match what happened. The record is public. The truth is verifiable. The headline is not the meeting.