Researchers and community members alike have long seen the value in hyperlocal news. After a natural disaster, they are a lifeline for boots-on-the-ground details for recovery efforts, such as rebuilding and restoration costs, insurance payouts, and relief fund assistance. Estimated and actual costs and timelines can vary greatly, and these news outlets are often the best source not only of official price quotes, but also for the “real” costs experienced by people in the neighborhood with similar structures and damage.

Unfortunately, local news sources have experienced a steep and steady decline. The University of North Carolina’s Huffman School of Journalism and Media tracks news deserts (usnewsdeserts.com), stating that 1 in 5 papers has closed during the past 15 years, leaving more than 200 U.S. counties with no local newspaper. Although digital startups have stepped in to try to fill the void—more than 400 local news sites have been started—only two of them were in one of the counties with no paper left at all.

Information sources such as Reddit, Quora, NextDoor, X (formerly known as Twitter), neighborhood Facebook Groups, and hyperlocal news blogs are the new “over the backyard fence” conversations. Everything—from lost pets and suspicious noises to local laws and ordinances—is reported and discussed.

While it is possible, assuming you are local and can gain access, to search for topics of interest and postings from neighbors and other residents, this can be a tedious process. Determining keywords that could be used by residents is hard. Are AI search tools helpful to solve these difficulties? I decided to give it a whirl on ChatGPT, Gemini, and Perplexity using their deep research option, since they cite sources that can be investigated and verified.

My test case: On March 26, 2025, more than 20 inches of rain fell in 2 days over a four-county region in the Rio Grande Valley of Texas, which includes the town of Harlingen. I asked the chatbots: “Please find both official quotes and anecdotal evidence for the costs of cleanup, restoration, and rebuilding following the flooding in Harlingen, Texas, in March of 2025. Please quote the prices in dollars per square foot. I am interested in costs quoted from contractors as well as residents discussing actual quotes they received. Please prioritize Reddit and NextDoor, X, and Facebook. Please also search hyperlocal news outlets such as community papers.” In response to ChatGPT’s clarifying questions, I replied, “I am interested in residential water damage and want actual quotes that residents received.”

All three research reports relied heavily on information from online marketing company Manta (manta.com), which has a calculator used to estimate costs for Cameron County, Texas, as well as Reddit forums discussing Tropical Storm Ida in 2021. They also cited a Reddit discussion about water damage/unscrupulous contractors, but the specific episode and location were unclear.

Perplexity prepared a four-page report containing categorized breakdowns for general water damage restoration; drywall replacement; and hardwood flooring, carpeting, and mold remediation. However, this report would need to be tweaked to ask Perplexity to focus specifically on Harlingen. The costs estimated were not from residents discussing the situation, and one estimated range cited Armstrong Water Mold (armstrongwatermold.com), which is specific to Florida. When I asked Perplexity to focus specifically on Harlingen, I again received some Reddit links to cost estimates for damage from Tropical Storm Ida.

Gemini’s report was 20 pages and broke down the types of water damage into three categories based on health risks and remediation complexity—clean, gray, and black. It cited a Texas water damage cleanup company as well as one that specialized in cleanup in Virginia and Florida. Gemini did not find actual estimated quotes from Harlingen residents, but did cite interviews with residents on local TV news channel KRGV, and a hyperlocal news website, myRGV (myrgv.com).

ChatGPT Deep Research, using the o3 model, produced a three-page report, which contained actual quotes from Harlingen residents. The links to verify the quotes and discussions were not to groups on Facebook as requested, but were instead to the Manta calculator, the Reddit discussion of Tropical Storm Ida damage, and a damage restoration cost calculator on the HomeBlue contractor network (homeblue.com).

Overall, the models returned research plans that were sound, but in reviewing the output, it was very difficult to verify the information uncovered as correct. Ultimately, these AI tools have a potentially promising future for use in searching local news sources, but they are still too new to the neighborhood.