How to find a domain for your startup

Paul Graham wrote 10 years ago that your startup ought to have a dot com domain. A lot has changed since then, but he still thinks there’s no excuse for not having one. PG is certainly more qualified to offer startup advice than I am, but a startup domain search is typically more nuanced than taking ten minutes to rebrand with a five-letter dot com.

Buying a good dot com

Let’s first figure out if you can buy a “good” dot com.

Do you have VC money? Some other source of money will do, but do you have millions of dollars? You don’t have to actually spend millions, but to spend tens or hundreds of thousands you should probably have millions. If you’re not ready to spend at least five figures, you likely can’t buy a “good” dot com, sorry—you can still buy an ok dot com.

Ok, so you have lots of money (nice!). Are you willing to rebrand? If your startup doesn’t have a name yet, that’s even easier—good on you for thinking about this now. But if not, are you willing to change your name? Your answer really changes the math for landing a good dot com. If you can’t rebrand, your dot com might be astronomically expensive or not for sale at any price.

Before you rebrand, you should of course reach out to the dot com owner of your startup name—either directly or through a broker—to at least get an asking price. I’ll cover tactics for contacting the seller and working with brokers further below.

But how much should you actually spend? I don’t have a precise percent-of-funding recommendation for your domain budget, but I get the sense that allocating anything above 10% of company funds is unusual. There are, of course, exceptions. AI companion company Friend spent $1.8 million—nearly three quarters of its funding—on friend.com. You could consult your investors so they aren’t blindsided by a big domain purchase, and they might even own some solid domains from failed portfolio companies. I’d like to compile more data on this as most domain sales are not made public, so if you’ve run a startup email me to share whatever figures you can.

At this point, if you firmly subscribe to the need-the-dot-com school of thought, buying alternatives could still be a good idea for defensive (read: security) or operational purposes. And likewise if you’re in the non-dot-com camp, you should still go through the exercise of searching for a dot com. You can own both and redirect one (e.g., vercel.dev redirects to vercel.com; valtown.com redirects to val.town).

Buying an ok dot com

If a “good” dot com is too expensive, can you buy an “ok” dot com? By ok I mean:

• Prefixed, e.g., try-, join-, use-, with-, go-, etc. For example, Ro started as getroman.com before moving to ro.co, and co-founder Rob Schutz has talked publicly about eventually buying ro.com (owned by a large Chinese company).
• Suffixed, e.g., -hq, -app, -co, -ai, or whatever makes semantic sense for your startup. Branding agencies Lexicon and Catchword use lexiconbranding.com and catchwordbranding.com, and it’s their job to name things well!
• Made up, obscure, or non-English words. For example, I just brainstormed for one minute and came up with remath.com, currently asking $2k aftermarket. There are also names where a startup-y suffix becomes part of the name itself, like Spotify.

Many people will argue that prefixed, suffixed, or made up/obscure dot coms are good and not just ok. Of course, many of these words have made it into everyday usage as verbs, like uber and google (not exactly a made up word but a typo). Many successful startups have made this work:

• Uber
• Lyft
• Airbnb
• Figma
• Tesla
• Reddit
• Canva

In other words, ok can become good, so you have an opportunity to get a bargain. By good I pretty much mean obvious value for many potential buyers.

If you want to avoid thinking about domains and focus on whatever it is your startup does, a dot com is simple. But if a “good” or “ok” dot com is too expensive, or you like the idea of an extension that fits your company branding, consider the wide world of alternative top-level domains (TLDs). The tradeoffs may be worth it, depending on your startup’s situation.

Buying an alternative TLD

As of September 2025, there are 1,592 top-level domains. That means there are 1,591 dot com alternatives, although some are restricted—they don’t give a .gov to just anyone. You have something like a thousand actual choices.

The first thing to understand is that a TLD is either a country code top-level domain (ccTLD) or a generic top-level domain (gTLD). Every two-letter domain extension is a ccTLD, and all the rest are gTLDs. As far as branding goes, this distinction often doesn’t matter. It is relevant for ccTLDs like .uk, .cn, and .de that are strongly associated with the country itself, but most of your customers won’t know or care that .ai is owned by the small island nation of Anguilla (and makes up about a quarter of their annual revenue!). But if you go with a ccTLD, you do need to trust that country’s commitment and competence.

ccTLDs

If you buy a ccTLD, you are trusting that country to keep the lights on, and many haven’t over the years. Some countries operate their own domain, like .co of Colombia. Others partner with a registry to operate the domain, like Anguilla does for .ai. In most cases you’ll be fine, but you should do some diligence before buying (or feel free to ask me for my two cents on a given ccTLD).

As a case study, consider .md. I think obsidian.md is the perfect domain for Obsidian (the note-taking app): files in Obsidian are markdown files like readme.md, so the domain mirrors that. It could also be a fitting extension for MDs—as in doctors—setting up a website or email for their private practice, for example. It turns out .md is the ccTLD of Moldova, the Eastern European country between Ukraine and Romania (slightly larger and more populous than New Hampshire, for reference). Moldova’s own government IT and cybersecurity organization operates the ccTLD and charges $40/yr and up for domains. Over 30k domains have been registered. After light diligence I’d say .md appears to be a stable ccTLD, although there are still other factors to consider.

Regularly used ccTLDs that are often not explicitly associated with the country (i.e., often used semantically for branding) include:

• .ai, Anguilla
• .co, Colombia
• .io, British Indian Ocean Territory
• .me, Montenegro
• .tv, Tuvalu
• .fm, Federated States of Micronesia
• .sh, Saint Helena
• .vc, Saint Vincent and the Grenadines
• .to, Tonga
• .ly, Libya

Some of these carry geopolitical risk (e.g., Libya), so do your homework. Or consider a gTLD...

gTLDs

Like ccTLDs, generics are run by registries. Unlike ccTLDs, registries control gTLDs by applying and bidding for rights in auctions (e.g., Google won the rights to .app for $25 million in 2015). Registries—whether run by a country or another organization—decide how much to charge registrars (where you buy domains) for a given TLD. In some cases ICANN—the governance organization that sets the rules for domains—is involved in pricing, too.

So the price you pay varies by TLD, and it can vary widely. Verisign charges $10.26 per .com, so you end up paying your registrar around $11 including the ICANN fee. On the pricier end, Google’s registry charges $400 for .new domains. Registries also mark some shorter, “good” domains as premium as they see fit.

Ok, you’d like to find a domain that fits your startup’s brand. There’s a good chance you have multiple options that make semantic sense, and you may even find an exact fit. For developer tools, there’s .dev. For e-commerce startups, .shop. There are occupationally-specific TLDs like .attorney, .dentist, and .doctor. Then there are broader options like .land, .page, and .live. There are even playful extensions like .fun, .cool, and .pizza. The list really is enormous, and it’s a fun exercise to scan through the options.

Once you climb back out of the gTLD rabbit hole, you have a handful of tradeoffs to consider before buying your domain.

Universal Acceptance

Universal Acceptance (UA) is an ICANN initiative calling for all domains to be treated consistently across the Internet.

UA is a concept or ideal, but in practice domains are not treated equally. More recent gTLDs and non-Latin TLDs are less likely to be treated as domains or valid email addresses by software. Implementation naturally lags behind, and the game of catch-up will continue with a big new round of gTLDs coming in 2026. Software assumptions around hyperlinks, web forms, email addresses, and the like will continue to go stale.

For example, I often notice that Apple Notes automatically creates a hyperlink for older TLDs like .com and .org but not for newer ones like .media and .press. To be fair, this is a bit of a UX challenge because auto-linking is not always what the user wants. When I send email newsletters for my domains book I often include domains that I really don’t want to be hyperlinked (like when I wrote that sex.com sold for $12 million—I don’t want readers to accidentally visit the site! Gmail also flags certain links as suspicious and might mark my emails as spam).

I’m eager to learn more about the UA effort at ICANN84 (ICANN’s annual meeting) next month in Dublin, and I’ll update this guide with any new learnings.

Customer perception

Capital U, capital A Universal Acceptance calls for computers to treat all TLDs equally, but lowercase u lowercase a universal acceptance (my distinction—not a recognized term) by humans is another question entirely.

Many domain experts I’ve talked to argue that the value of dot com is only rising with the sea of alternatives. And many friends I talk to outside of the tech bubble say they’re suspicious of newfangled TLDs. I hear the same arguments over and over: dot com signals legitimacy, longevity, credibility. It’s a compelling case.

Customer perception is highly contextual, of course. If your startup caters to grandparents, then a shiny new TLD may cause confusion. If you’re shipping a devtool, .dev, .run, .new, .build, or .sh could work.

If a domain is cheap, I’d go ahead and buy it, but I also recommend sleeping on it and asking around for a bit to uncover unknowns. You might find problems, like: customers might think .co is a typo for .com; customers might be confused if another company owns the dot com for the name you’re buying; or it might fail the radio test.

The “radio test” asks whether a customer will type a domain in correctly after hearing it verbally. I’d say airbnb.com doesn’t quite pass the radio test—I bet many people would’ve typed something like airbandb.com when they first heard it. My book’s website, dotcom.press, does not pass the radio test—it’s kind of a hassle, to be honest, but I’ve been happy with the tradeoffs so far.

Email sender reputation

Email clients like Gmail do not explicitly factor TLD into your domain’s sender reputation, but certain TLDs do have more of a reputation for spam.

It’s a good idea to review reputation data by gTLD and ccTLD on Spamhaus. They cover domains, malware, botnets, phishing, and spam for both gTLDs and ccTLDs. For example, .info, .club, and .email are all in the gTLD top 10 for spam reputation, which is a bummer because I like all three for the right website or startup (like the email service buttondown.email, which switched over to buttondown.com for $85k last year). TLD reputation alone isn’t reason to rule out a domain, but it’s worth considering.

When it comes to email, you should take concrete steps like setting SPF, DKIM, and DMARC records wherever you handle DNS. You can also verify your domain(s) with Google to decrease your chances of landing in spam. Subdomains accrue separate sender reputations, so you might want to divide emails by function using subdomains (e.g., one for marketing, another for billing).

SEO

Like email sender reputation, SEO does not explicitly consider TLDs, but it can indirectly influence organic search traffic. Factors like perceived credibility can impact click-through rate.

If you’re buying a domain on the aftermarket, you should consider whether it was used before. A site with lots of historical traffic could be an SEO asset (which is why domains for popular sites will be squatted if they expire). On the other hand, a domain with a malicious track record could harm your SEO or email sender reputation.

Btw, if you disagree with any of this or have relevant information to add, email me and I’ll update accordingly. This applies to the entire guide, but especially for email and SEO where I’d like to collect more data.

Tactics

Registrars

Trust is important in the domain world. You need to trust the registrar you use to search and buy domains. Some registrars engage in shady behavior like purchasing domains for their own portfolio based on your query data, for example.

To buy an available domain in 2025, you don’t have to pay a markup above what the registrar paid the registry. Vercel and Cloudflare both offer at-cost domain registration. Vercel just launched a snappy new domains interface last week and has been rolling out support for more and more TLDs. Instant Domain Search is another gold standard domain search UX for available and aftermarket domains, and they link out to partner registrars where you can actually purchase.

There are too many registrars to list, but I’ve asked dozens of startup founders and developers which registrar they prefer, and the ones I hear (in a positive light) most often are: name.com, porkbun, namecheap, iwantmyname.com, and of course Cloudflare and Vercel. Amazon also has a registrar called Route 53 buried in AWS, and Google shut down Google Domains a couple years ago. If you’re a happy camper at a registrar I missed or an angry customer at one I included, let me know and I can update this guide.

Contacting the seller

When you have your eyes set on a parked domain, tread lightly. Depending on where it’s listed, the seller could be watching your every move to gauge interest and gain negotiating leverage.

For starters, if you plan to engage with a broker, you shouldn’t reach out to the seller yourself. The last thing you want is the seller to think they have competing offers. Your broker can clear this up after the fact (“that was my client who emailed you last week”), but it’s probably better to proceed with caution. Visiting the parked domain’s website over and over again is another act of signaling interest to the seller. Play it cool. Instead of visiting the website directly, you can use the Wayback machine to see what’s been there over time.

If you have a major funding announcement coming up, hold off until you have the domain in hand. One founder told me they waited to announce getting into YC until after their five-figure dot com purchase was finalized.

You could consider paying for a domain via lease-to-own, and if your startup continues to go well you own the domain at the end of the lease. I’ve also heard of startups offering equity in exchange for a domain, but I get the sense that’s somewhat rare. Dharmesh Shah recently sold os.ai to Perplexity and chat.com to OpenAI for equity, so it happens.

If you do want to contact the seller directly, you’d previously use a WHOIS record, but contact information is almost always private nowadays (since GDPR), so you may have no choice but to contact the seller through whatever parking spot they’ve set up. I’m not an experienced negotiator, so consider my advice here as a starting point—I’ll leave the rest to brokers.

Domain brokers

As with registrars, there are many domain brokers out there with varying histories and incentive structures. Some focus on buy-side work with startups, others deal with domain investors more often. Many have their own inventory of domains. Some charge a fee up front for their services, and others wait to charge until a transaction goes through. Some consult with naming, others only track down specific domains.

As with registrars, trust is key, so I get the sense that brokers tend to lean on referrals. Your investors and founder friends are a good place to find a broker you trust. In my startup and Twitter circles I often hear Snagged and Lumis come up, for example. If you’ve had a good (or bad) experience with a broker, email me.

What else?

I’d like this guide to be a living document, so please reply below or via email with things you disagree with, anecdotes on your startup’s domain, questions, or anything else on topic. Thanks for reading.