Americans are increasingly turning to artificial intelligence to help decipher the thorny and ever-changing labyrinth that is the U.S. healthcare system. Three in five Americans say they recently used AI tools for health or healthcare queries, according to a survey published Monday by ChatGPT creator OpenAI. The survey was part of a larger report tracking similar trends globally. The report acknowledged difficulties with dealing with U.S. healthcare requirements are likely an important driver of this user trend.

“In the United States, the healthcare system is a long-standing and worsening pain point for many,” the report read. Tracking between 1.5 and 2 million health insurance-related messages each week, OpenAI found Americans use ChatGPT for everything from seeking diagnoses to comparing coverage plans and dealing with post-visit bills and claims. 

Even before ACA subsidies expired, the complexity and high cost of interacting with the healthcare system was discouraging for many Americans. Around 27 million Americans are uninsured, according to the CDC, with many more underinsured, meaning they have coverage but it rarely pays out enough to cover the cost of a claim. A December Gallup poll found a record low 16% of Americans were satisfied with the healthcare system. 

The U.S. healthcare system is bogged down by an abundance of stakeholders, which researchers say has contributed to higher administrative costs. These include private plans offered through employers, federal providers such as Medicare and Medicaid, and plans offered through ACA, which offered income-based subsidies for individual coverage on federally run marketplaces. 

In 2023, nearly 60% of insured Americans said they experienced problems using their health insurance, according to a survey by health policy research outfit KFF. That survey also found around nine in 10 insured adults would support policies to simplify healthcare, including requiring insurers to maintain up-to-date provider directories and give clearer disclosures about out-of-pocket cost liabilities. 

The system’s complexity was already funneling many patients to AI tools like ChatGPT for help, the OpenAI report found. The company also claimed people living in communities with low coverage rates could stand to benefit the most from having real-time access to healthcare information. On average, nearly 600,000 healthcare-related queries are submitted to ChatGPT every week from so-called “hospital deserts,” underserved rural areas at least a 30-minute drive from the nearest general medical center.

The trouble with using ChatGPT as your medical advisor

AI can be helpful in deciphering bureaucratic bloat, though researchers have warned of the risks when users become too reliant on AI for medical diagnoses. More than 60% of Americans say AI-generated health information is at least somewhat reliable, according to a survey last year by the University of Pennsylvania. But AI medical diagnoses can also suffer from inaccuracies and bias, as studies have found large language models can issue vastly different recommendations based on a patient’s race, income level, and sexual orientation.

Issues also arise when users turn to AI for mental health advice. Last year, OpenAI was hit with seven lawsuits in California alleging ChatGPT use had contributed to multiple suicides and psychological harm. In November, OpenAI tweaked its usage policies to walk back ChatGPT’s ability to give medical advice. 

OpenAI did not immediately return Fortune’s request for comment. In a statement to Business Insider related to the usage policy change, an OpenAI spokesperson said: “ChatGPT has never been a substitute for professional legal or medical advice, but it will continue to be a great resource to help people understand legal and health information.”

With millions suddenly lacking any coverage at all in the wake of ACA’s subsidy expiry, many more Americans could tap that resource soon.