Source: 2026 AI x Journalism Summit
In May, members of Consumer Reports’ Experimental Engineering Team traveled to Baltimore for the AI x Journalism Summit. We spent two days with speakers and fellow attendees from newsrooms, academia, and industry wrestling with how to use AI without sacrificing trust. The summit was a great chance for us to learn from other organizations whose work depends on building trust and fact-checking.
Here are some highlights and takeaways:
Future-Proofing Facts
The workshop Future-Proofing Facts opened with a description of the current state of journalism: fake AI freelancers fooling major outlets, reading lists citing books that do not exist, synthetic “experts” quoted in published pieces, and newsrooms’ own AI tools inventing quotes in summaries.
Hilke Schellmann and Loreben Tuquero described a layered process for verification that included provenance standards like C2PA, reverse image search, source callbacks, document-level quote verification, and AI detection.
Smart, Confident, and Wrong: Designing Responsible A.I. Tools in the Newsroom
In the New York Times’ Smart, Confident, and Wrong session, Dylan Freedman and James O’Toole demonstrated how AI assistants misrepresent content a large fraction of the time. Their work to “make invalid states unrepresentable” involves constraining the tool’s interface and outputs so a hallucinated or unsupported claim has no way to reach the page looking like a verified one. Citations, confidence, and source links are all structural parts of the output.
Freedman distinguishes between “internal” tools (with the reporter-as-filter and lower external stakes) and “reader-facing” tools. Reader-facing tools get held to a higher bar: their outputs are constrained more tightly, they must be red-teamed before shipping, and the audience cannot be left to do the verification the tool should have done itself.
Large Language Mathematicians: Public Records in Record Time
Tyson Bird’s Large Language Mathematicians session presented American City Business Journals’ work surfacing stories from public records at scale. In their early attempts, they let the model “do the math” and it hallucinated numbers. They eventually landed on a system that uses AI to infer schema and help shape messy inputs, runs deterministic code for arithmetic and validation, and reserves the LLM for tasks where fuzziness is acceptable.
Building an AI Tool That Finds News Amid the Noise
Peter Rasmussen walked through the Baltimore Banner’s News Detector which monitors 100+ Maryland sources, scores stories on journalistic signals, and gives editors a prioritized list of leads. The Banner highlighted the need to define what “good” means with stakeholders, measure false positives and misses, and resist treating speed as a substitute for rigor.
Listening at Scale: Building AI Tools for Audio and Video Monitoring
Kaveh Waddell’s (a Consumer Reports alum!) session Listening at Scale showcased how AI systems can help journalists listen at scale. Kaveh demoed a real-time police scanner monitor and RoganBot, a system analyzing how narratives spread on The Joe Rogan Experience.
Spreadsheets, Not Chatbots
Aaron Brezel’s (a former fellow at Consumer Reports!) session showed how, for investigative and research work, the most useful interface may be a spreadsheet. The idea of “Spreadsheet Inference” is that large, messy reporting questions can be made manageable by sorting information into rows and columns, letting formulas and repeatable steps handle the grunt work.
Instead of asking one big prompt to “find the answer,” reporters use AI to comb through documents, extract signals, and organize patterns in a structure that remains visible and auditable at every phase. Using a spreadsheet preserves provenance: the sources, extracted fields, and formulas applied.
Aaron’s gas-ssi-toolkit is very cool and I’m working on ways to incorporate it into our manual evaluation flows for our RAG chatbot AskCR.
If you are working on verification, source quality, or consumer-authorized AI and want to compare notes, reach out at innovationlab@cr.consumer.org.
