Short answer: it can, if your notes list conditions you did not evaluate or treat. Blue Cross Blue Shield companies just measured that pattern in hospital claims: an estimated $942 million in added costs from 2023 to 2025, about 70 percent of it from secondary diagnoses, while treatment did not rise to match. Keep every diagnosis tied to something you did in the visit.

Clinicians have been warning each other about this in their own forums: a scribe that writes down everything it hears can put a condition on the claim that the visit never touched. This month, the insurers put a number on it.

What did Blue Cross find?

On September 24, the Blue Cross Blue Shield Association published an analysis of hospital claims. It estimates $942 million in added spending for Blue Cross and Blue Shield companies between 2023 and 2025, driven by patients being coded as more complex. About 70 percent of it, more than $650 million, came from secondary diagnoses: conditions listed alongside the main reason for the stay that moved the claim into a higher-paying category.

The association says more than 60 percent of hospital systems now use AI tools that scan notes, lab results and records for diagnoses that affect coding. Reuters reported that the association pointed to two routes: software that combs existing records, and ambient scribes that listen to the visit and draft the note.

Its evidence is the gap between the codes and the care. The hospitals it studied recorded significantly more anemia diagnoses without a matching rise in transfusions. Luke Chalker, the association's senior vice president of product and data science, told Reuters the disconnect suggests that

“AI is identifying more billable conditions, not sicker patients.”

What do hospitals say?

That patients really are sicker. The American Hospital Association disputes the finding and argues that patient acuity has genuinely increased, citing demographic shifts and better documentation. An analysis by the association and Vizient found that case mix index, a standard measure of how sick hospital patients are, rose about 5 percent from 2019 to 2024.

Neither side has published patient-level chart reviews that would settle it. For a practice owner, the argument matters less than the direction: payers are now measuring what AI adds to the note.

Does this apply to a private practice?

The analysis looked at hospital stays, not office visits. The rule for office visits is older and plainer. The official ICD-10-CM coding guidelines say to code the conditions that coexist at the visit and “require or affect patient care, treatment or management,” and not to code conditions that were treated before and no longer exist.

A scribe cannot tell which is which. The patient mentions the diabetes another doctor manages, the knee that healed last year, a lab value from last month, and the scribe can write all of it down. Whatever codes the visit then reads what the scribe wrote. One commenter on a health IT news site named the worry: using software to “mine charts for more billable diagnoses.”

How do I keep my notes out of trouble?

1. Read the diagnosis list before you sign. Every diagnosis should connect to a line in your plan.

2. Say what you did out loud, so the note records the care and not just the conversation.

3. Switch off code suggestions you do not use, and check every code you keep.

4. Keep resolved conditions as history, not as active problems.

5. Once a month, pull a week of claims and look for diagnoses that are new since you started using a scribe.

Frequently asked questions

Did the Blue Cross analysis look at office visits?

No. It looked at hospital claims for Blue Cross and Blue Shield companies from 2023 to 2025. The coding rule for office visits points the same way: code what required or affected the care you gave.

Do AI scribes cause upcoding?

Not by themselves. A scribe writes down what it hears. The risk comes when every condition it hears becomes a coded diagnosis without a clinician checking it against the care that was given.

Does Solanto add diagnoses to the note?

Solanto will not. It will write the history from the visit, leave the assessment and the diagnoses to you, and keep anything it suggests out of the note until you put it there.

The short version

Insurers are now measuring what AI adds to the note. Keep every diagnosis tied to something you did, and read the list before you sign. Solanto will document, not code: it will write the history, leave the judgment to you, and run in Cloud mode or in Private Mode, where the note never leaves your device.

The price is on the page, not behind a call. Early access is open: solanto.ai/pricing

Send this to the colleague whose notes doubled in length after they started using a scribe. Ask them to read the diagnosis list.