With AI Help, More People Appeal Their Property Tax Assessment

A chatbot offering homeowners personalized advice boosted tax challenges, which are concentrated among the more well-to-do

Every year, the Dallas County appraisal authority sends out proposed property tax bills to more than half a million homeowners. These notices also state that the homeowner can appeal, seeking a lower bill, if they think the authority has overvalued their property. Roughly 24% of them did earlier in 2026, either independently or by contracting with a professional to do it. 

In a working paper, researchers ask why that appeal rate isn’t higher. Dallas County property tax rates are substantial, creating an average bill of $7,923 in 2026, and appeals are free, the working paper points out. There are no filing fees or no-show penalties, and no professional service required. Appeals, or “protests” in Texas terms, result in higher tax bills only in very unusual situations.

Some of the team’s previous research showed that about 70% of county homeowners who filed protests in 2020 reduced their bills, for an average savings of $485. From that, one could surmise that 2026 Dallas homeowners spent tens of millions of dollars more in property taxes than necessary. 

Within Dallas County’s quite diverse homeowner population, the researchers note, Black, Hispanic and nondegreed homeowners and those who own homes valued below the county median are the least likely to capture these savings, largely because they protest less often than other demographics. For example, the 2025 appeal rate was 41.7% among homes valued above $513,000; 9.9% among homes below $265,000.

In the study, offering the county’s homeowners an AI chatbot to help prepare a 2026 property tax appeal substantially increased the likelihood that homeowners generally would file a protest, according to the authors, University of Michigan’s Justin E. Holz, University of Virginia’s Andrew Simon, UCLA Anderson’s Ricardo Perez-Truglia and University of Texas at Dallas’ Alejandro Zentner.

But the tool may have widened existing demographic disparities between those who do and don’t benefit from the appeals process, the researchers suggest, with caveats. Visitors across demographics engaged with the bot, but it’s unclear whether these interactions increased filing rates among disadvantaged demographic groups, the researchers write. They note with confidence a significant, statistically solid rise in the rate of appeals among nonminority, college-educated and higher value homeowners who used the bot.

Care to Chat?

Dallas County is huge, with a politically and demographically diverse population larger than 15 U.S. states, the study notes. Property taxes are the key source of revenue for local services, as Texas does not collect an income tax. Its property tax appeals process is uniform across the state and similar to models common in other parts of the county. Generally, taxing authorities treat protests as a necessary factor in keeping valuations accurate year after year.

The researchers created a website that provided step-by-step instructions on how to file an appeal, as well as personalized property information the study subjects would need to do so. A participant could, for example, type in their address and download a “comparable properties report” detailing homes similar to their own. While they could not submit the report itself, they could pick examples from the report and paste those details into their filing to support a claim. 

For about half of the visitors, the site also offered an AI chatbot, powered by Anthropic’s Claude, to answer questions about filing a tax appeal for the owner’s particular property.

About two weeks before the April 14 protest filing deadline, the researchers mailed postcards to 45,200 households inviting them to the website. By design, the sample excluded homeowners who had already hired paid professionals to appeal. The invitation did not disclose that there were two versions of the site. 

Of the 645 who responded — likely those already considering a protest — about half were assigned to the site with the bot. The study results compare the rate of protest filings and other responses in visitors who saw the chatbot to those who did not. 

From Bot to Action

Some 78% of study participants who had chatbot access engaged with it. Popular queries included asking advice on how to decide if they should file an appeal and which comparable sales they should use to build a case. In the chatbot responders group, 50.5% filed an appeal. Just over 41% of those who visited the bot-less version filed.

Another 9.5% of overall visitors went on to hire an agent to file an appeal. Bot interactions didn’t seem to affect these decisions. Subjects assigned to the chatbot group hired professionals at roughly the same rate as those who reviewed the data and process without bot help. Among all Dallas County homeowners who filed protests this year, about three-quarters hired an agency to file.  

If a chatbot can give more homeowners confidence in filing appeals on their own (for free), the technology could, theoretically, shrink the differences in benefits from appeals across demographics. In the study, demographic groups that historically file fewer tax appeals engaged with the bot at only slightly lower levels than the other cohorts.

But the increases in filing rates in these subcategories appeared to be much smaller, and possibly, statistically insignificant, the researchers note. “Together, these results provide suggestive evidence that universal access to the chatbot may, far from mitigating existing inequities, exacerbate them,” they write.

It’s very likely that a lot of the control group in this experiment — those at the no-bot site — also used other AI to prepare protests, the authors note. The experiment could only measure how filings changed when subjects were offered a free AI tool specifically trained to help with their property tax appeal. While this particular chatbot spoke to taxpayers across demographics, it did not spur them to act equally.