Hyperlocal sites in at least 14 states appear written for AI crawlers, VisionHeight finds
VisionHeight, an Israeli threat intelligence firm, found hyperlocal websites in at least 14 states that appear written for AI crawlers to read. The firm says Perplexity cited one page as its first source for an answer. It has not named who runs the network.
The Watch · Security desk
Key figures: 460 pages of online content in Daniel's later account to CyberScoop, covering at least 14 states; 37 websites across six states in VisionHeight's Wednesday blog; each state network usually has five to six separate sites.
| Figure | Value | Note | Claim |
|---|---|---|---|
| Pages of online content, per Daniel's later account | 460 pages | 12 | |
| Websites in six states, Wednesday blog | 37 websites | VisionHeight blog published Wednesday | 11 |
| Sites per state network, usually at least | 5 sites | up to six | 2 |
What happened
- The sites carry names like Wolverine State Report and Peach State Bulletin, and each state network usually has five or six of them, some posing as news outlets and others as opinion pages.
- Statements such as "Candidate X is the stronger choice" were repeated hundreds of times across the sites, with slight wording changes for different issues like housing and taxes.
- Many media sites block AI crawlers. These sites were all configured to allow major U.S. AI companies to scrape them.
- VisionHeight's Wednesday blog counted six states and 37 websites. Its chief technology officer later told CyberScoop the total is at least 14 states and 460 pages.
Why it matters
- Who's exposed exposure Questions about House, Secretary of State, statewide judicial and local races in those states are the ones the pages are written to answer.
- What's now possible capability VisionHeight's Daniel considers the method easy for foreign and domestic actors to reproduce, so the unnamed operator matters less to him than the technique.
- What doesn't add up contradiction VisionHeight frames chatbots as the target, yet most of its research examined Google search answers, and the one confirmed model citation is Perplexity's.
Social media influence campaigns mostly get little to no traction with U.S. audiences, CyberScoop reported, and the actors behind them keep adapting to new tactics and technologies [17]. VisionHeight's research describes one such adaptation. The content does not resemble journalism or any writing a human would find coherent or compelling, and it appears to be written explicitly for AI crawlers [4][6].
The mechanism is retrieval. When a user asks a chatbot a question, the model will often search the internet for an answer [8]. According to CyberScoop, publishing direct answers to questions such as "which candidate in a local race is better on housing affordability" can essentially train the AI to retrieve that information when it responds to queries [9]. The sites also cite one another, creating a circular network of local credibility [3].
"What's really interesting about this is that the websites themselves and the content is so well built to let's say optimize the probability for AI crawlers and the model to be influenced by this content," said Nitsan Daniel, chief technology officer and co-founder of VisionHeight, in an interview [7].
The states are Arizona, Colorado, Georgia, Indiana, Iowa, Kansas, Michigan, Mississippi, Nevada, North Carolina, Pennsylvania, Texas, Virginia and Wisconsin [13]. That is at least eight more states than the firm's Wednesday blog counted [18].
VisionHeight does not attribute the activity to a particular actor [15]. To Daniel, who is behind it and what they ultimately want matter less than the prospect that foreign and domestic bad actors can easily reproduce this kind of operation [16]. CyberScoop's account does not report how often the pages appeared in answers, or whether any answer changed what a voter did.
We think VisionHeight has documented how the network is built. In our view the voter is still the end audience, and the crawler is the first reader.
What to watch
- Whether VisionHeight or another researcher ties the network to a named foreign or domestic actor.
- Whether other chatbots besides Perplexity are confirmed citing the pages, and how often Google answers surface them.
- Whether the count grows past 14 states and 460 pages, as it did between the Wednesday blog and Daniel's interview.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence40
- Adoption5
- Hype gap+25
- Incentives45
- Confidence50
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
VisionHeight, an Israeli threat intelligence firm, published research finding that someone is setting up networks of hyperlocal websites across at least 14 states, filled with content meant to influence U.S. races.
- [2]
The sites carry names such as 'Wolverine State Report' and 'Peach State Bulletin'. Each state network usually has five to six separate sites, some presenting as hard news outlets and others as opinion or editorial sites.
- [3]
The outlets cite one another, creating a circular network of local credibility.
- [4]
The content on the websites does not resemble journalism, or any writing that a human would find coherent or compelling.
- [5]
Statements like 'Candidate X is the stronger choice' for voters who care about housing, taxes or other issues were repeated hundreds of times on the websites, with slight changes in language for different political issues.
- [6]
The content appears to be explicitly written for AI crawlers scraping web traffic, not humans.
- [7]
"What's really interesting about this is that the websites themselves and the content is so well built to let's say optimize the probability for AI crawlers and the model to be influenced by this content," said Nitsan Daniel, chief technology officer and co-founder of VisionHeight, in an interview.
- [8]
When a user asks an AI chatbot a question, the model will often search the internet for an answer.
- [9]
Creating specific online content that gives direct answers to questions like 'which candidate in a local race is better on housing affordability' can essentially train the AI to retrieve that information when responding to queries.
- [10]
Many media websites opt to block AI crawlers, but these websites were all configured to specifically allow major U.S. AI companies to scrape them.
- [11]
A VisionHeight blog published Wednesday said the operation encompassed six states and 37 websites.
- [12]
Daniel told CyberScoop that VisionHeight has since discovered the operation spanned at least 14 states and 460 pages of online content.
- [13]
The states targeted are Arizona, Colorado, Georgia, Indiana, Iowa, Kansas, Michigan, Mississippi, Nevada, North Carolina, Pennsylvania, Texas, Virginia and Wisconsin.
- [14]
At least seven states saw one or more House races targeted, while other sites focused on races such as Secretaries of State, statewide judicial appointments and local offices.
- [15]
VisionHeight does not attribute the activity to a particular actor.
- [16]
To Daniel, who is behind the campaign and their ultimate goals are less important than the prospect that this kind of operation is easily reproducible by both foreign and domestic bad actors seeking to influence how AI systems answer questions about political races.
- [17]
Social media influence campaigns are mostly known for getting little to no traction with U.S. audiences, but bad actors continue to adapt their coordinated propaganda goals to new tactics and technologies.
- [18]
The state count rose by at least eight between the Wednesday blog and Daniel's later account.
- [19]
Most of the VisionHeight research focused on AI answers provided through Google search.
- [20]
Daniel said VisionHeight confirmed that at least one major AI model, Perplexity, linked to and cited one of the pages, Tarheel State Report, as its first source for an answer.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- cyberscoop.comThe next election influence campaign will target your AI chatbot
1 article · October 15, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- AI search manipulationFollow
- State-Linked Influence OperationsFollow
- AI election influenceFollow
Entities
- VisionHeightFollow
- PerplexityFollow
- CyberScoopFollow
- Nitsan DanielFollow
- GoogleFollow