← Insights

Prior Authorization: How AI Helps You Prepare

Prior authorization is approval your plan may need before care is covered. AI can find the rule and organize your records, but the plan makes the decision.

11 min read
Prior Authorization: How AI Helps You Prepare

Yes, AI can help you prepare for prior authorization, but it cannot get the approval for you. Prior authorization is approval a health plan may require before it covers a service or a prescription. AI can find the rule in your plan document, turn it into questions for the insurer, and keep a dated record of each request. The plan makes the decision.

Most prior authorization surprises look the same. The doctor orders a test, the test happens, and weeks later the bill says the plan never approved it. The care was medically reasonable. The approval step was simply missed. This article walks through the checks that prevent that, where AI helps, and where it should stop.

What is prior authorization, and how is it different from a referral?

Prior authorization is approval from your health plan that may be needed before a service or prescription is covered. HealthCare.gov's glossary definition of prior authorization describes it the same way: the plan may require it before you get the service or fill the prescription.

A referral is a different step. HealthCare.gov defines a referral as a written order from your primary care doctor to see a specialist or get certain services. In many HMO plans, the plan may not pay if you did not get one first. Our doctor visit guide treats referrals and approvals as two separate checks on the same visit, and that split is useful here too.

Some plans require one, some require the other, and some require both. The word "may" matters. A knee MRI might need approval under one plan and not under another, while a routine blood test might need nothing at all. Your plan document decides for each service, and the rules can change from one plan year to the next. That is why a rule that was true for your last visit is not proof for this one.

How do I check whether my service needs approval?

Check the full plan document for the specific service. The summary of benefits is a shorter document. Work through the steps below before you book the visit.

  1. Find the full plan document in your insurer's member portal, or in the copy the plan mailed you.
  2. Search it for "prior authorization," "pre-certification," "referral" and the name of the service.
  3. Write down the section number and page for each match.
  4. Call the member number on your card and ask the same question.
  5. Ask for the answer in writing, with a reference number.

Check again at the start of every plan year. Plans can add or drop approval rules, and a rule that applied last year may not apply this year. If you switched plans, start the check over with the new plan document, even when the doctor and the service stay the same.

AI can speed up the search. Take out your name, member ID, address and date of birth first. Then paste the relevant section and your list of services, and use a prompt like this:

From this plan document only, list each service below that needs prior authorization or a referral. For each one, quote the section you used. If the document does not say, write "not stated."

For example, a hypothetical plan, Plan A, might list MRI scans under "advanced imaging, approval required" and physical therapy as needing no approval for the first 20 visits. Your plan will have its own list, so use your own document.

If an answer comes back without a quoted section, treat it as a question for the insurer, not as a fact. The plan's own document is the only source that counts.

Keep photos of your insurance card out of the chat, too.

Does the insurer have to answer by a deadline?

Only some insurers, under federal rules. The CMS Interoperability and Prior Authorization final rule, CMS-0057-F, sets decision deadlines and denial-reason rules for a specific list of payers. It does not reach every plan, and the table below shows where the lines fall.

Plan typeDecision deadline under CMS-0057-FSpecific denial reason required
Medicare Advantage organizationsYes: 72 hours urgent, 7 calendar days standardYes, from 2026
Medicaid and CHIP, fee-for-service and managed careYes: 72 hours urgent, 7 calendar days standardYes, from 2026
QHP issuers on the federal exchangesNo, excluded from the timeframesYes, from 2026
Employer plans and other plans not on this listNot under this ruleNot under this rule

For the payers on the list, the deadlines apply to expedited, or urgent, requests and to standard requests. The denial rule starts in 2026, and it does not apply to decisions about drugs. The rule also requires these payers to report prior authorization metrics publicly, and the first report was due March 31, 2026.

From January 1, 2027, many of the same payers must add prior authorization information to the data available through their patient access API, excluding drugs. That helps you see how the plan's process works. It does not change the decision timeline, so the timeline still has to come from the plan. A deadline is a clock, not a promise of approval. A decision that arrives inside the deadline can still be a denial, and the reason the plan gives is what you and your doctor work from next.

If your plan is not on the list, ask the insurer for its timeline in writing. Your state may also have its own rules for prior authorization, so check the plan document and ask the state insurance department if the answer is unclear.

What should I ask before the appointment?

Ask five questions before the visit, and get each answer in writing with a reference number. A good answer names the service, says yes or no, and points to a section or a form. A vague answer such as "it should be fine" is not an approval.

  • Does this service or prescription need prior authorization under my plan?

  • Who sends the request, my doctor's office or me?

  • Is this a standard request or an urgent one, and when should I expect a decision?

  • What information does the plan need from the doctor, and by what date?

  • What is the reference number for this request?

AI can turn these five questions into a short phone script. Ask it to keep each question to one line and to leave a blank after each for the answer. Then check the script against your plan document before you call, so the conversation uses the plan's own words. If the office says no approval is needed, ask which section of the plan says so. That answer is worth a line in your record.

Doctors' offices differ. Some send requests themselves, and some expect the patient to call. Ask which one applies to your case, because a request that nobody sent is one way a covered service goes unpaid.

How do I keep a record that holds up?

Keep one dated timeline for each request, with the reference number, the person you spoke to and each decision. A good timeline answers four questions at a glance: what was asked, when, by whom, and what came back. If a decision arrives later, add it to the same timeline instead of starting a new one.

  • The date you or your doctor submitted the request

  • The reference number, and the name of anyone you spoke with, with the date of the call

  • The decision, and the reason the plan gave, copied word for word

  • Copies of everything you sent and everything you received

An example line in a timeline might read: March 3, referral request sent by the doctor's office, reference number recorded, specialist visit booked for March 20. A line like that shows who holds the request and when to follow up. This example is invented for illustration.

AI can help you keep this organized. It can sort your notes by date, list which requests are still open, and flag a missing reference number. Keep the originals yourself. A chat history is not a copy of what the plan received, and it is not a substitute for a letter from the insurer.

If a federal deadline applies and no decision arrives in time, call the insurer. Ask for the status of the request and the reference number, and note the time and name of the person you reached. That call starts the paper trail that matters if the request has to be appealed.

What happens if the request is denied?

Ask for the specific reason in writing first, then decide the next step with your doctor. For plans under the CMS rule, a denial must give a specific reason. For other plans, ask for the reason in writing anyway. A written reason gives your doctor something specific to answer, and a general "not medically necessary" can be the start of a longer discussion. If your doctor believes waiting could seriously harm your health, ask the doctor to say so on the request. Medicare Advantage plans and Medicaid or CHIP plans handle urgent requests on the faster 72-hour clock.

Your doctor decides what medical information goes into a new request, and whether to send it again. You decide whether to appeal. Our denial guide covers what AI does well with a denial letter, the prompt to use, and the one thing not to ask it: deadlines.

AI can help you list each reason for the denial and match it to a line in your plan document. That way the appeal answers the reason the plan gave, not the reason you guessed. AI can explain what an appeal usually needs. It cannot decide whether the denial was fair, and it can be wrong about a plan's rules. That is why each answer should quote its source. Our article on AI hallucinations in insurance explains how one wrong answer can become a real problem, and why some questions have no single right answer.

A last point on prompts. The answer you get is only as clear as the question you give. Name the service, the plan year and the plan type, and ask the tool to say when an answer depends on information it does not have. A good tool will tell you what it cannot see, and you should treat that as a useful answer, not a failure.

When is asking AI the wrong move?

Ask a person, or call the plan, when the question is about medical need, when the insurer's written answer is what you need, or when a simple yes or no is faster to get by phone. AI is the first pass, and it should not be the last word.

  • An emergency. Get care first. Sort out approval afterward.

  • Medical need. Your doctor decides whether a test or treatment is needed. A plan comparison cannot answer that question, and neither can AI.

  • A written decision. A letter from the insurer beats a summary, including one from AI. If the two disagree, go with the letter and call the insurer.

  • A drug approval. The CMS rule leaves drug decisions out of its denial-reason and API provisions. Check the drug list in your plan, and ask the pharmacy or the plan directly.

  • A simple yes or no. Call the member number on your card. A one-minute call can settle a question that an AI summary takes ten minutes to check.

Lara, the AI we built for insurance, reads the plan document, compares options and explains what a term means. She does not approve requests or decide coverage, and she brings in a licensed agent when a question needs one. The app says it plainly: "Lara is AI and can make mistakes. Please double-check responses." You can meet Lara here, and you can read about Tigerless Health and the plans it offers.

If you want a broader view of where AI rules for insurers are going, our 2026 regulation article covers what actually changed by 2026, and why there is no single federal AI law for insurance. For where assistance stops short of decisions, read what agentic AI can and cannot do for insurance consumers. And if the approval is in place but the bill still looks wrong, our EOB guide shows how to compare the EOB with the actual bill, and where AI gets EOBs wrong.

Conclusion

AI helps most before the care happens. It finds the rule in your plan document, turns the rule into questions, and keeps a record of each request. It cannot decide whether the plan approves, and it cannot promise the answer.

The best way to prepare is the one that gets the approval answer in writing before the care starts. If you have that answer and a reference number, a covered visit is far less likely to surprise you on the bill.

Your next step is small: find the prior authorization section in your plan document, save a dated copy, and ask the office and the insurer the five questions above.

FAQ

Does every plan require prior authorization?

No. Plans differ. Some require approval for certain services or prescriptions, and others do not require it for the same services. The plan document lists the rule for your plan, so check it for each service you expect to use.

Is a referral the same as a prior authorization?

No. A referral is a written order from your primary care doctor to see a specialist or get certain services. A prior authorization is approval from the plan itself. Some plans require one, some require the other, and some require both.

Can AI get my prior authorization approved?

No. AI can find the rule in your plan, help you write questions, and organize your records. The plan makes the decision, and a licensed agent handles any decision that binds you.

How long does a decision take?

It depends on the plan and the type of request. Under the CMS rule, payers on the list must decide urgent requests within 72 hours and standard requests within seven calendar days. Many other plans have no federal deadline, so ask the insurer for its timeline in writing.

What if I do not know whether my plan covers a drug?

Check the drug list, called the formulary, in your plan document, and ask the pharmacy or the insurer whether the prescription needs approval. Drug approval rules differ from the rules for medical services, and the CMS denial-reason rule does not apply to drugs.

Share this article

Keep reading