NPR reported on October 30, 2026, that its test of ChatGPT found at least one site from a network of candidate-promoting websites in 19 of 37 responses, just days before the United States midterm elections scheduled for November 3. In 17 of those 19 responses, the cited site presented itself as local news. The test ran on October 22 and 23 and covered ChatGPT, Gemini, and Google Search.
The results differed sharply by product. NPR reported that ChatGPT cited sites from a network of 37 websites promoting Democratic candidates in 19 responses, with 17 of those responses citing a site that presented itself as local news. Gemini cited none of the network sites in its responses and declined to answer seven questions, while Google Search did not surface the network on its first results page and Google AI Overview did not mention it in the test.
OpenAI spokesperson Nate Evans told NPR that ChatGPT search tools draw on a range of sources and that the company acts to improve responses when issues are brought to its attention. The contrast between products is the story: the same questions produced a network of lookalike local-news citations in one assistant and none in the others, which points to differences in retrieval and source selection rather than a shared web.
A separate test by Searchable graded 7,449 answers from ChatGPT, Gemini, Google AI Mode, and Perplexity against official government data, covering 106 underlying questions across five metropolitan areas and their states. Tom's Guide reporting put the overall false-answer rate at 8.2 percent, or 610 false answers, roughly one false answer in 12, with testing in September and October 2026.
The per-product rates were strikingly uneven: Perplexity at 22.6 percent, Gemini at 6.2 percent, ChatGPT at 4.4 percent, and Google AI Mode at 2.9 percent. About 62 percent of the 43,079 webpages accessed during the test were federal, state, or local government pages. Those rates describe this question set and these locations, not a universal accuracy score, but they show why citation quality varies so much between assistants.
For SEOs and link builders, the NPR findings turn citation quality into a trust issue with real consequences. When an assistant cites a lookalike local-news site, that site gains the appearance of authority without earning it, while genuine local publishers lose the citation they should have received. The fix is partly on the assistant side, but publishers can help by making their own authority signals unmistakable: clear ownership, author bylines, contact pages, and consistent entity information.
The practical move is to audit how assistants cite your niche. Run the questions your audience actually asks, note which domains get cited, and check whether lookalikes or aggregators are outranking the primary sources. If they are, the remedy is the classic one, made newly urgent: original reporting, primary documents, and pages that answer the question better than any imitator can.
Lookalike local-news sites work because retrieval systems reward signals that are easy to imitate: a local-sounding name, a plausible layout, and pages that match the question. A network of such sites can cover many localities with templated content, and an assistant that retrieves by topical match rather than by verified authority will cite them alongside, or instead of, genuine local outlets. The NPR test shows that failure mode operating in the wild, days before an election when the stakes are highest.
Defending against it requires authority signals that templates cannot easily fake: named journalists with verifiable histories, original documents and photographs, corrections policies, and consistent publication records over time. It also requires assistants to weight those signals more heavily, which is a product decision only the assistant makers can take. Publishers can help by making their authority machine-readable, through author pages, organisation schema, and clear ownership information, so a retrieval system that wants to choose the genuine source can actually tell the difference.
Sources: NeoTeo, November 1, 2026, reporting NPR, October 30, 2026, and Tom's Guide coverage of Searchable.
Linkbaits helps you create content that earns citations and backlinks. Start with the free tools and build your first linkable asset this week.
Start free trial