The Radio Test: How to Tell if Your Domain Name Is Actually Memorable (2026 Guide)

You're on a podcast. The host asks, casually, at minute 38, "So where can people find you?" You say your domain. The host repeats it. The listener — who is driving, with a coffee in one hand — has roughly four seconds to either remember it or forget it forever.

That four-second gap is the Radio Test, and in 2026 it's harder to pass than ever. Podcasts, voice search, AI assistants reading citations aloud, TikTok founders mentioning their site in a 30-second clip — your domain name lives more often as spoken sound than as a typed URL. If it doesn't survive being said once, it doesn't exist.

This post breaks the Radio Test into 4 sub-tests you can run on any candidate in about 60 seconds, plus the cognitive-science reason each one matters and the specific failure patterns that keep killing otherwise-good names.

Why "memorable" is harder to define than founders think

Most founders test memorability by saying the name to a co-founder, a partner, or a friend who already knows the company exists. Those people will remember it. They're cheating: they have context, they want you to succeed, and they've already heard the pitch.

The right test isn't "can someone who likes me recall this?" — it's "can a distracted stranger who has never heard of us reconstruct this name from sound alone when they sit down at a laptop tomorrow?" That's a much higher bar, and most domains fail it for one of four predictable reasons.

The 4 sub-tests

Test 1 — The Spelling Test phonetic

Say the name out loud. Could someone who has never seen it written spell it correctly on the first try?

This sounds trivial. It isn't. The graveyard of failed domains is full of names that look clever in writing but fall apart spoken aloud:

• Lyft works because the misspelling is famous and consistent. Lyftr would not.
• Fiverr survived because the double-r is regular enough that the brain accepts one canonical typo.
• Quizzlr, Snapply, Klyk — all variations of the same trap: the founder thinks dropped vowels are quirky, the listener types four wrong things and gives up.

The rule: at most one non-obvious letter. Two is a coin flip. Three is a lost user.

Test 2 — The Echo Test distinct

Does the name sound like something that already exists? If your candidate is one syllable off from a competitor, a famous brand, or a common word, listeners will round it to the more familiar thing and end up on the wrong site.

Try this: say your candidate ten times in a row. Then ask someone to repeat it back without seeing it written. If they say something close-but-different — Notably instead of Notibly, Stride instead of Strydd — you've got an echo problem. The brain rounds unfamiliar sounds to familiar ones; that's not a flaw to engineer around, it's how memory works.

This is the single most common reason "available .com" names underperform. The .com was available because the name is almost something else, and search traffic leaks to the something-else.

Test 3 — The Boundary Test parseable

Two-word domains are some of the strongest brands of the last decade (Stripe is one word but SendGrid, MailChimp, and SquareSpace are not). They work because the boundary between the words is unambiguous.

The Boundary Test: when you say the name, is there exactly one obvious place to put the slash between syllables? If the answer is "two possible splits," you've got a domain that requires the listener to also do a small linguistics puzzle while remembering it.

Examples:

• PenPath — clear boundary, one parse. pass
• Therapie — could be ther-apie or thera-pie. slow
• Experts.exchange — three competing breaks. fail

If a stranger needs to hear it twice to find the seam between words, your domain costs you one in three listeners.

Test 4 — The Recall Test sticky

The first three tests check whether someone can reconstruct the name correctly. The Recall Test checks whether they'll still have it in memory 12 hours later — when they finally sit down at a laptop.

Run it like this: say your candidate to a friend in passing, talk about something else for at least an hour, then ask them to type it into a browser. If they can't, the name doesn't have what cognitive scientists call distinctiveness — the property of being unusual enough to leave a residue in long-term memory.

The cheapest distinctiveness booster in 2026 is a concrete image. Names that evoke a physical object (Cobra, Anvil, Loom) outperform names that evoke an abstraction (Synth, Quanta, Nexus) by a meaningful margin on 24-hour recall. The brain stores images more durably than concepts.

What a perfect score looks like

A name that passes all four tests has these traits:

Test	Pass signal	Common failure
Spelling	At most one non-obvious letter; phonetic English	Dropped vowels, silent letters, cute substitutions
Echo	Doesn't rhyme with a bigger brand or common word	"It's like X but with a Y" sounds like X
Boundary	One unambiguous syllable break	Two-word compounds that mash at the seam
Recall	Anchored to a concrete image or strong consonant	Abstract Latin or Greek roots

Five names that pass — and one that almost did

To make the framework concrete:

• Stripe.com — one syllable, English, concrete image, no echo. Perfect score.
• Notion.so — phonetic, distinct, single boundary. The .so requires saying "dot S O" but the root word is unambiguous.
• Linear.app — common word, no spelling ambiguity, the suffix doubles as a category cue.
• Loom.com — one syllable, concrete object, distinct from every other software brand.
• Vercel.com — invented but obeys English phonetics (sounds like "parcel"), low echo risk, easy to spell on first hearing.

The almost: Xero. Phonetically identical to "zero" — so on the radio it works, but typing it requires knowing the X. Built a billion-dollar company anyway, but spent years training their market to use the X. Most founders do not have years.

How to actually run the test in 2026

You don't need a focus group. Three steps:

1. Record yourself saying the name three times, naturally, like you would on a podcast. Listen back. If you have to clarify or spell it on the recording, that's a failure.
2. Send the recording (audio only) to five strangers — Discord, X DMs, a Slack group. Ask them to type back what they heard. Anything below 80% match rate kills the candidate.
3. Re-test 24 hours later with the same five people. Don't remind them. Whoever still types it correctly is your true memorability rate.

If you're shortlisting more than 5 candidates, generate them first and run the test only on the survivors of basic availability and length checks. For that, you want a generator that filters to phonetic, distinct, image-anchored names with available TLDs in one pass.

Where founders most often fail

Two patterns account for roughly 80% of Radio Test failures we see across the names founders share for review:

1. Optimizing for "available" before "memorable." The cheapest unclaimed .com is almost always cheap for a reason. Spell it out: it's a misspelling, a four-syllable Latin word, or a homograph. Memorability comes first. Availability is a constraint, not a goal.

2. Trusting yourself. You are not a normal listener. You have already heard the name a thousand times in your head. Every name sounds memorable to its author. The only honest test involves people who do not know you and are not paid to be polite.

One last thing: even a perfect name doesn't replace the basics. Run a trademark check, confirm the social handles are also available, and verify the name doesn't translate to something embarrassing in your second-largest target market. The Radio Test is necessary but not sufficient.

If you've already got a shortlist and want to pressure-test it against pivots, brandability, or SEO drag, the related guides below walk through each angle in detail.