How to Name an AI Healthcare Startup in 2026

1. Why your name decides 5% of your fundraising odds

Digital health startups raised roughly $4 billion across 110 deals in the first quarter of 2026, and 59% of that capital landed in just twelve mega-rounds of $100M or more — Whoop's $575M Series G at a $10.1B valuation, OpenEvidence's $250M Series D, Verily's $300M, Talkiatry's $210M Series D, AcuityMD's $80M Series C, Courier Health's $50M Series B and a small handful of others (eMarketer, Q1 2026 health-tech funding analysis).

That distribution is the entire story. A founder competing for any of the 98 sub-mega deals is fighting a triage problem, not a quality problem. Partners screen 200 inbound decks a quarter. The ones that get a meeting are the ones that survive the first thirty-second skim — and a name lives at the top of every page in that deck, every Crunchbase pull, every PitchBook entry, every cap-table reference, every press hit. A name that adds friction at any one of those steps subtracts from the only resource that decides outcomes: partner attention.

I am not claiming that a better name causes a $200M Series D. I am claiming something narrower and more useful: you cannot afford a name that adds friction at every step from cap table to PR. If your firm's name needs to be spelled twice on a phone call, mispronounced by a senior analyst, or explained to a clinical advisor before they will introduce you, you have already lost the compounding effect that a clean name produces over a 24-month fundraise.

This guide is for founders before they spend $5,000 on a domain or $50,000 on a rebrand. It is the framework I use when I evaluate startups, and the same framework I apply to the domains I curate at MedNameHub.

2. The four name archetypes in 2026 healthcare AI

Almost every defensible name in healthcare AI today falls into one of four archetypes. They are not equal — each signals something different to clinicians, payers, and investors, and choosing the wrong archetype for your category is a quiet but compounding mistake.

The contrast pairs are where the choice gets sharp. A mechanism name like ctdna-anything credentials you with oncologists but means nothing to a hospital CFO. A specialty name like precision psychiatry compresses your pitch into two words for a payer but boxes you in if you later expand into neurology. A stance name like clearprior signals taste and an opinion about where care is going, which is exactly what Series A partners price; it also leaves room for the inevitable category-renaming that happens around any new technology.

Outcome names — dischargeflow, readmitless, codecorrect — are the most underrated of the four. They are boring to a brand consultant but legible to a hospital COO at a glance, which is the only audience that matters for an enterprise SaaS sale. If you are a B2B workflow product, default to outcome unless you have a specific reason not to.

3. The 7-second test: would a patient understand this name?

Read your candidate name aloud to a 70-year-old patient. Then, in seven seconds, explain what the company does. If you cannot, the name is wrong — not because patients are your buyer (most AI healthcare startups are B2B), but because clinicians internally use the patient-comprehension test as a proxy for whether a name is serious or marketing-flavored.

"ClearPrior" passes. A daughter sitting next to her father can be told "the software makes the insurance approval part of the visit transparent," and the seven seconds are intact. The name does the work; the explanation just confirms it.

"Recursion" — the publicly traded drug-discovery company — fails the test in isolation. The explanation "an AI platform that finds new drugs" requires you to disconnect the meaning of the word from what the company actually does. Recursion has the market cap and the publication record to carry that name; a seed-stage company does not. The lesson is not that abstract names fail — it is that abstract names require multiples of brand investment to anchor, and that capital comes out of your runway.

The reason this matters more in healthcare than in fintech or consumer software: clinicians are the gatekeepers to every workflow you want to enter. A name that sounds like it came from a Silicon Valley naming workshop will not survive the first lunch-and-learn at a large health system. Buy clarity early, or pay for it later in clinician-onboarding cost.

4. The TLD decision tree: .com, .ai, or .health?

This is the part of the naming problem that is worst-served by the open web. Most TLD advice was written for direct-to-consumer e-commerce in 2017. Healthcare AI in 2026 has a different decision tree, and it should be walked in this order:

1. Are you B2C, FDA-regulated, or selling a device or drug? Use .com. Regulators, journalists, and patients default-type .com when they search for you. Trust costs more than originality at this end of the market.
2. Are you a developer-tools company, an inference platform, or an agentic-ops product where your buyer is an engineering team? Use .ai. The TLD is now a category signal — a way to preempt your position before competitors do. Engineering buyers read it as deliberate.
3. Are you selling to payers, policy stakeholders, or research-tools customers? Use .health. The TLD specializes you instantly and saves a sentence in every cold email. Payers and academic medical centers respond well to it because it does not pretend to be a consumer brand.
4. Are you a health-system B2B SaaS company? Default to .com. Only fall back to .tech if the .com you want is unaffordable, and even then expect a small but real signal-loss with traditional health-system buyers.

The cleanest test of this tree is the contrast between two names in the same portfolio. clearprior.ai was the right call even though clearprior.com would have looked "cleaner" on a business card. The buyer for a prior-authorization AI product is a CMIO or revenue-cycle leader with an engineering team behind her; the .ai signals that the founder respects the technical stack. clinicalaiscribe.com is the inverse — a workflow product where the buyer is a CMO or COO running an RFP. The .com reads as institutionally serious to that buyer, and it survives the second-page printout of a procurement memo.

A practical mapping of TLD trust in healthcare 2026:

• .com — universal default. Lowest cognitive overhead for any audience. The price you pay for one-word .coms is the price of removing friction.
• .ai — developer respect, premium positioning. Read by engineering and clinical-AI buyers as a category claim, not a fallback.
• .health — regulator and payer positioning. Strong with academic medical centers, weak with patients.
• .tech — usable but under-resourced. Read by some VCs as a signal that the founder could not afford the .com. Avoid unless your runway demands it.

One nuance that founders miss: if you take .ai, buy the matching .com defensively even if you do not plan to use it. The cost of the .com is almost always less than the cost of a competitor (or a typo) catching your traffic two years in. xAI learned this; so did Character.AI. Both pay attention to their .com posture for a reason.

5. Trademark screening in 60 minutes (USPTO + EUIPO walkthrough)

You cannot get a clean trademark opinion without a lawyer, and you should not try. What you can do in an hour is rule out names that obviously will not survive a $5,000 IP clearance memo six months from now. That hour is the highest-leverage time in the entire naming process, because every dollar you spend after this — domain purchase, logo work, decks, registration — is at risk of being thrown away if the name does not clear.

Walk through these five steps in order:

1. USPTO TESS search. Open the USPTO trademark search system and run your name, then run all phonetically similar variants. If your candidate is Augmentyx, you must also check Augmentix, Augmentics, Augmentex, and Augmenticks. Examiners use phonetic equivalence as a likelihood-of-confusion factor; you should too.
2. Confirm the right International Classes. For an AI healthcare startup, you almost always need to check Class 9 (downloadable software), Class 42 (SaaS, technical research, and software development), and Class 44 (medical services). Add Class 10 if you sell devices, and Class 5 if your name touches pharmaceuticals. The WIPO Nice classification defines what each class covers.
3. EUIPO eSearch. Repeat the same query at the EUIPO eSearch Plus database. Do not skip this even if you have no near-term EU plans. A European competitor with a five-year head start on your mark is a problem at Series B, not a problem at Series A — but it is still your problem.
4. Common-law search. Google your name + healthcare and + software. Look for unregistered users who have been in the market for five-plus years. Common-law rights are not visible in TESS but are enforceable, and a startup using your candidate as a product name (even without a registration) can block your application or force a coexistence agreement.
5. Apply knock-out criteria. If you find an active registration in Class 9 or 42 in a healthcare context, stop. Do not try to argue you are "different enough." Examiner-level likelihood-of-confusion analysis is granular and unforgiving, and your differentiation argument is what your lawyer will pitch — at $400 an hour — only after you have already paid for everything else.

One pattern worth noting from TESS data: combinations of augmented and therapy are dense with active registrations in Class 9 and Class 44, across mental health, physical therapy, and digital therapeutics applicants. Anyone considering an "AugmentedTherapy"-style name should expect knock-out hits in the first ten minutes of a TESS pass. The fix is not to differentiate the modifier — it is to switch archetypes entirely (see chapter 2): move from a modifier+modifier construction to a mechanism or outcome name that does not share vocabulary with a crowded class.

6. The clinical-credibility test

This is the chapter only a clinician can write, and the one that separates a name that closes hospital contracts from a name that wins a design award.

The test: imagine your company name printed on the bedside EHR screen of an ICU. Can the nurses read it without smirking? Would a radiologist write it into a dictated report without rewording? Would a hospital CMIO put it on a slide for the board without first scrubbing the marketing tone out of it?

The anti-pattern in 2026 is clearer than in any year before. Names like MagicMed, FastDx, QuickCare, and HealEasy read to clinicians as commerce talking past evidence — the same register as supplement marketing, weight-loss programs, and direct-to-consumer pharmacy ads. Hospitals filter these aggressively in vendor procurement because the legal and PR exposure of a poorly-named partner is real. Your name is, in part, a procurement-risk variable.

The pro-pattern uses real clinical vocabulary. ctdnahealth.com uses ctDNA — circulating tumor DNA — a term every oncologist recognizes from the liquid biopsy literature. The name credentials itself before the first call. virtualnursingai.com uses virtual nursing, which is the term of art in post-acute and inpatient care for tele-supported bedside nursing models. The Agency for Healthcare Research and Quality and major health systems use the phrase verbatim in published work; the name slots into a CMIO deck with no rewording.

The one question to ask before locking a name: when a doctor Googles this for the first time, does it sound like a serious new company or like someone selling vitamins? If the answer is uncertain, the name is wrong.

7. Pricing reality: what you should actually pay

The premium-domain market is opaque on purpose, and most published pricing guides are written by sellers. Here is the unvarnished version of where prices actually clear in 2026 for healthcare-AI names, based on observed transactions across Sedo, Afternic, and Atom.

Realistic ranges by type:

• One-word generic .com in a medical or clinical domain (medical, clinical, oncology, diagnostic): $30,000–$200,000+ on premium marketplaces. These rarely make sense for a pre-Series A founder unless the word is the entire brand thesis.
• Two-word .com in a modifier+modifier or modifier+noun construction: $1,500–$8,000. This is the sweet spot for most founders, and where the highest ratio of brandability per dollar lives.
• .ai compound (two words): $800–$3,500 for unowned names; $5,000–$15,000 for names with category traction.
• .health compound: $300–$1,500. The TLD's pricing has not caught up with its rising signal value.
• .tech compound: $200–$800. Often a sign that someone could not afford the .com, which is itself a problem for institutional buyers.

The practical sweet spot for most pre-seed and seed founders is $1,500 to $6,000. In that band you can typically buy a name that is premium-listed, trademark-pre-screened, and brand-able without burning your domain budget. Below it, you are usually buying something that should have been listed on a discount marketplace; above it, you are paying for a level of memorability that pre-product-market-fit companies cannot yet monetize.

On marketplaces: Sedo has the widest curated inventory and the most reliable escrow process. Afternic wins on distribution because of its GoDaddy integration, which means the same name often shows the same price across dozens of registrars. Atom (formerly Squadhelp) is the most curated and the most expensive per ranking — its strength is brand-able made-up words, its weakness is that the price reflects the curation premium rather than the secondary-market clearing price.

When to negotiate: Make an Offer listings are almost always negotiable, typically 20–40% below the asking number, especially if the domain has been listed for more than twelve months. Sellers respond to the first reasonable offer because the alternative is paying renewal fees for another year.

When to pay buy-it-now: when the name is the entire company business card and a 20% discount is not worth a six-week delay. Calculate the rebrand math honestly. By the time a seed-stage company has signed its first twenty enterprise contracts, a $5,000 BIN price is a rounding error against the avoided cost of relaunching the brand. A $50,000 rebrand at Series A includes the domain, new logo work, updated decks, contract addenda, integration documentation, app-store updates, and the soft cost of customer confusion — none of which are visible from where you stand at seed.

8. Six examples from a working portfolio

The following six names are on MedNameHub right now. They are listed not as endorsements of any single startup category, but as worked examples of the framework above — each one is paired with the archetype it fits, the TLD logic behind it, and the buyer it speaks to.

These are six of 209 names currently on MedNameHub. The right one for any given founder is the one that survives chapters 3, 5, and 6 of this guide — which is to say, the one a clinician can read aloud, a USPTO examiner cannot knock out, and a CMIO will say without rewording.

9. Three traps that have killed perfectly good names

The three patterns below come up repeatedly in pre-seed naming reviews. Each is a category of mistake, not a single name — and each is easier to see in someone else's startup than in your own.

Trap 1: The "Augmented Therapy" archetype

USPTO TESS shows dense registration of augmented + therapy combinations across Class 9 and Class 42 — physical therapy, mental health apps, digital therapeutics, and AI-rehab tools have all converged on the same vocabulary. If you want to communicate that you are doing AI-augmented therapy, the modifier+modifier construction is the wrong tool. Switch to a mechanism archetype (name the underlying technique) or an outcome archetype (name what the patient experiences). A composite scenario: a Series A team I reviewed had selected a name in this pattern, paid $4,800 for the .com, and discovered at the IP-clearance step that two senior registrants in Class 9 covered nearly identical mark constructions. The cost of the lesson was the domain plus six weeks of brand work.

Trap 2: The "AI + Health" monogamma

The literal words AI and Health appear together in thousands of registered company names — a quick LinkedIn search returns more than 2,000 matches and rising. If your company name is some variant of AIHealth, HealthAI, or HealAI, your brand has effectively zero distinctiveness. Every advertising dollar you spend is also paying to disambiguate yourself from companies with the same name shape. The fix is a modifier that does the differentiation: ConsentHealth survives a TESS scan and a Google search; AIHealth dies in both. The point is not the specific modifier — it is that one of the two words has to do real semantic work.

Trap 3: Medical + verb names that read as treatment claims

MedHeal, CureFlow, PainGone, HealFast — these read to a clinician as treatment claims, and they read to the FDA the same way. The agency monitors brand and product names for implied therapeutic claims, particularly for drugs, biologics, and devices that go through review. FDA guidance on proprietary names explicitly addresses overpromising name constructions. If you are an early-stage company that has accidentally named yourself in a way that implies a therapeutic outcome you have not proven, expect to be asked to change it at the regulatory-review stage. A name change at FDA review is not a logistical inconvenience — it is a six-month brand crisis, and it ripples into every contract, integration, and clinical-trial document you have signed under the old name.

10. The post-launch checklist

The domain is the visible piece, but it is not the whole brand surface. Before you announce the company — or, ideally, before you take it out of stealth — confirm the following are in place:

• Defensive TLDs purchased. If your brand is on .ai, own the .com as well. The reverse also applies. Renewal cost on a defensive TLD is rounding error compared to losing a typo-traffic competitor at Series B.
• LinkedIn company handle claimed in the exact brand spelling. Five minutes of work that saves five weeks of stealth-naming awkwardness during a Series A diligence sweep.
• X (Twitter) handle claimed. Even if you do not plan to post for a year, the handle becomes a credentialing artifact when reporters search for you.
• WHOIS privacy enabled if you are in stealth. Default registrar settings on Porkbun, Cloudflare, and Namecheap enable this; double-check on transfer.
• Catch-all email aliases live. sales@, hello@, founders@, security@. The last one matters more than founders realize — researchers and bug-bounty hunters will email security@ before they email anything else, and a bounce is a credibility hit.
• GitHub organization claimed in the brand name, even if you have no public repos. Engineering hires check.
• USPTO and EUIPO trademark applications filed before you send your Series A pitch deck. A pending application is a diligence-friendly signal; the absence of one is not.

The domain is one of seven things you own. None of the other six are expensive, and all of them prevent specific, named failure modes downstream.

11. Closing: what would I do today?

If I were starting an AI healthcare company tomorrow, here is exactly how I would spend the first 90 minutes on the name:

1. 15 minutes: list ten candidate names, two or three from each archetype in chapter 2.
2. 15 minutes: run each through the seven-second test in chapter 3. Cut the half that do not survive.
3. 30 minutes: walk the remaining five through the USPTO TESS knock-out in chapter 5. Cut anything with an active registration in Class 9 or Class 42 in healthcare.
4. 15 minutes: score the survivors on the clinical-credibility test in chapter 6.
5. 15 minutes: open Sedo and check .com plus .ai availability for the top two finalists. If both are available and trademark-clean, you have a name. If only one is available, decide on the TLD using chapter 4.

If you want to skip to a shortlist that has already passed steps 1, 3, and 5, the MedNameHub catalog has 209 names organized by specialty and TLD. Either way, the point is the same: the framework is what protects you from the $50,000 rebrand. The name is the easy part once the framework is honest.