Choosing a domain name is often the first step in establishing an online presence for a new project or team. This article outlines a simple process for generating candidates, evaluating them for memorability and technical constraints, and selecting a final option that suits your specific needs without relying on complex SEO tactics.

Defining Your Project's Core Identity

This is a hypothetical scenario, and the author below is invented to show how the opening step might look in practice.

Imagine a two-person team planning a small tool for tracking reading lists, working under a working title of "Shelf Notes." Before opening any search box, they write a one-sentence identity statement: "A calm, no-frills reading tracker for people juggling three books at once." From that sentence they pull out three anchor words: shelf, stack, and quiet. They also decide what the name must not suggest—no library-software feel, no speed or productivity language. That negative list matters as much as the positive one, because it rules out candidates early instead of after an hour of arguing.

They then build a short identity sheet with four fields they can reuse later: audience (casual readers), tone (calm, plain), anchor words (shelf, stack, quiet), and exclusions (no "lib," no "fast"). Two made-up example rows show how they'd classify candidates:

Candidate: "quietstack" — anchors met: quiet + stack; tone fit: high; exclusion check: passes; note: reads more like a tool than a service.

Candidate: "shelfloop" — anchors met: shelf; tone fit: medium; exclusion check: passes; note: "loop" hints at returning to unfinished books, which matches the use case.

They settle on a rule for the first round: every candidate must contain at least one anchor word or a close variant, and none may contain an excluded word. When one teammate proposes "PaperShelf," they keep it even though it lacks quiet, because it fits audience and tone and the rule allows a single anchor. A different proposal, "ReadFast," gets cut immediately—it passes the anchor test only loosely and fails the exclusion list outright. Writing down that reasoning in the sheet, rather than just deleting the idea, gives the pair a record of why some directions were dropped.

The point of this exercise is to produce a filter, not a winner. By the end of the sheet, the team has a tone, two or three anchor words, and a written exclusion list they can hold every later candidate against.

Generating Simple and Short Candidates

Once the identity sheet exists, you can turn anchor words into tiny root words before adding anything else. A useful private move is to cut each anchor to one syllable or four to six letters, then write three variants by hand: one intact, one shortened, one with a single consonant changed. For the reading-list example, “shelf” might yield shelf, shel, and skelf; “quiet” might yield quiet, quie, and quet. These fragments are not names yet, but they give you a smaller alphabet to recombine. You might keep them in a column beside the anchor so the link stays visible.

From those fragments, you can form short combinations and test them against three quick questions on the same sheet: does it read cleanly when spoken once, does it stay under roughly ten characters, and does it avoid the excluded words already recorded? A hypothetical pass might produce “quetshelf” (too long, spoken once: unclear), “shelq” (short, but the final q feels abrupt), and “skelf” (short, distinct, but the consonant change drifts from the anchor). You can mark each with pass, revise, or drop, and write one-line reasons rather than scores.

Consider keeping a second column for alternatives you almost chose. If “skelf” is dropped for drifting, note “stef” as a closer variant; if “shelq” is revised, note “shelk” or “shelqo.” This preserves your reasoning without forcing a decision, and it keeps the shortlist honest about why each candidate survived. You can also group candidates by construction pattern—intact anchor, clipped anchor, blended pair—so later rounds do not quietly favor one pattern.

Before moving on, you might read the surviving candidates aloud in a single sentence, such as “I bookmarked it at shelk,” and note any that make you pause. Pauses are useful information, not failures. The aim at this stage is a small, explained set of short forms you can carry into the next comparison, not a final choice.

Checking Character Limits and Technical Constraints

For internationalized domain names, the source states that DNS stores these names as ASCII strings using Punycode transcription. The 63-octet label limit and the 255-octet domain name limit apply to the Punycode transcription of the labels, not to the original Unicode characters. The source also lists limits that TLD registrars can place on IDN naming conventions: restrictions on the characters that can be used, whether characters from more than one script can be used in a label, which characters can appear next to each other, and the length of each label.1

The source notes that IANA coordinates how domain names and IP addresses are managed, while TLD registrars control the rules and requirements for registering second-level domains. It gives the example that in the United Kingdom, .co.uk is used for commercial organizations while .ac.uk is used for academic institutions. These statements describe the source's stated scope; they do not establish additional limits beyond those the source names.1

Take one surviving candidate from your shortlist, such as “shelk,” and write it on a single line with its intended ending beside it, like “shelk plus a project-site ending.” You can then check that pairing against a few practical limits you set for yourself before any search: count the letters in the name alone, count them again with the ending attached, and note whether the spoken form still sounds like one word rather than two.

If your first choice runs long, you might write a trimmed variant directly underneath it and keep both visible. For example, “quietstack” could sit above “quista,” with a short note beside each saying how many letters it uses and whether the shortened form keeps at least one anchor word recognizable. You can decide which line to keep based on what still reads cleanly when said once, rather than on length alone.

A small table works for this: candidate, name length, ending used, combined length, and a yes-or-no column for “easy to say once.” Fill one row with “shelk,” a project ending, the combined count, and your own spoken check. Whether you keep that row or replace it with a shorter entry is your call; the table simply makes the trade-off visible before you settle on a favorite.

Avoiding Confusing Homoglyphs and Typos

Unicode Technical Standard #39 describes how visually similar characters can create security problems when an identifier can be misread as another entity. Its example is a user who needs to be sure they are interacting with a particular financial service rather than a spoofing entity. The standard notes that machines distinguish any two character sequences easily, but humans may struggle, so restricting which characters are allowed in identifiers is a security technique.2

The standard's Restricted set covers characters not in modern use, those used only in specialized fields such as liturgical characters, phonetic letters and mathematical letter-like symbols, and characters in limited use by very small communities. It applies canonical equivalence when testing candidates, so a candidate passes if either the base letter or the combining mark is Allowed, or if the precomposed form is Allowed. These checks belong to the Unicode standard, not to a domain registrar or naming tool.2

Take one surviving candidate from your shortlist, such as “shelk,” and put it beside every ending you are seriously weighing. Read each combined form once slowly, letter by letter, then once as a word. You might keep a narrow column for endings that turn the name into something you would misread aloud, like a lowercase l next to an uppercase I in a handwritten note, and mark those as drop-for-now rather than deleting them.

A practical private check is to write the candidate in a plain sentence twice, once in all lowercase and once with only the first letter capitalized. If “shelk” still reads as “shelk” to you both times, it passes this step; if you catch yourself hesitating between “shelk” and “sheik,” consider setting it aside. This is your own eye and ear doing the work, not a test result.

You can also compare lookalike pairs as a short written note: “rn versus m,” “0 versus o,” “1 versus l.” Pick two pairs that worry you most for the names you are keeping, and record which candidate each pair affects. A candidate that survives both readings might move forward; one that needs an explanation each time probably should not.

Before you settle, ask a teammate to spell the name back to you after hearing it once, and note the spelling they produce next to yours. Agreement is a useful signal; a confident wrong spelling is a reason to reconsider the ending, not the anchor word.

Evaluating Keywords and Brand Fit

When weighing keywords, Google's SEO guidance says to do what is best for the business when picking a site name, since users will use that name to find you, and to follow general marketing best practices. From a ranking perspective, the guidance states that keywords in the domain name or URL path alone have hardly any effect beyond appearing in breadcrumbs.3

The document also notes that ideas about keyword use have changed over time, and it lists keyword stuffing, repeating the same words or variations excessively, as against Google's spam policies and tiring for users. The guidance adds that Google Search does not use the keywords meta tag.3

For this step, you can take one surviving candidate, such as “shelk,” and add a keyword note to the same identity sheet: which anchor word it echoes, whether that echo is obvious or faint, and whether the faintness suits the project. “Shelk” echoes “shelf” only loosely, which may read as a quiet nod rather than a label; if the team wants the reading-list connection to be clearer, “shelfa” or “booksta” could sit beside it with a note saying why each is more or less direct. You might mark each row as clear, subtle, or missing so the pattern of your choices becomes visible.

Brand fit is a separate question from keyword fit, and you can judge it by rereading your tone line before each candidate. If the tone is “calm, plain,” then “shelk” may fit while an energetic coinage such as “shelzo” may feel off, even if both contain the anchor. Write a short reason beside every candidate you keep, as in “shelk — plain, one syllable, fits the tone; shelzo — playful, clashes with calm.” These notes are for your own sorting, not a verdict; a candidate that fails tone today may return if the project’s voice changes.

You can also decide in advance which matters more when the two pull apart: a keyword that is instantly readable, or a brand feel that is slightly less literal. Record that preference in one line, such as “clarity before cleverness,” then apply it to your rows. The remaining candidates are your choices to weigh, and the written reasons will make the final comparison less a matter of mood.

Finalizing and Recording Your Choice

Once you have a shortlist, you can turn your working notes into a single decision record you will actually keep. Take the two or three candidates that survived your read-aloud and spelling checks, and give each one a row with fields you choose yourself: candidate, the anchor word it echoes, the ending you paired with it, the combined form, your tone-fit note, and a one-line reason to keep or drop it. A filled example might read: candidate “shelk,” anchor “shelf,” ending “project suffix,” combined “shelk plus suffix,” tone fit “calm, plain,” verdict “keep — short, easy to say once.” Leave the verdict blank until you have compared every row side by side.

A decision is easier to defend later if you record the runner-up as well. You might add a column for “close alternative and why it lost,” then fill it with something like “quist — trimmed, but the dropped vowel made the spoken form less obvious.” This is not busywork; it is the note that stops you from reopening the same debate in a week.

Before you treat a name as final, write the exact string you intend to use, in lowercase, on its own line, and read it once more with fresh eyes. Consider whether the ending changes how the root reads, and whether the combined form still matches the tone line you wrote at the start. If it does not, you can move to the next row rather than forcing a fit.

You might also set a stopping rule in advance: for instance, “I will choose the first candidate that passes the spoken test and keeps the tone note intact.” Write that rule on the same sheet so the choice is made by your criteria rather than by mood. When a row wins, mark it clearly, note the date you decided, and keep the full sheet with your project notes. That single page tells you later what you weighed, what you rejected, and why — which is often more useful than the name alone.

Frequently asked questions

What fields can I put on a short identity sheet for later reuse?

They then build a short identity sheet with four fields they can reuse later: audience (casual readers), tone (calm, plain), anchor words (shelf, stack, quiet), and exclusions (no "lib," no "fast").

How can I turn anchor words into tiny root words before adding anything else?

Once the identity sheet exists, you can turn anchor words into tiny root words before adding anything else. A useful private move is to cut each anchor to one syllable or four to six letters, then write three variants by hand: one intact, one shortened, one with a single consonant changed.

What can I record in a second column for alternatives I almost chose?

Consider keeping a second column for alternatives you almost chose. If “skelf” is dropped for drifting, note “stef” as a closer variant; if “shelq” is revised, note “shelk” or “shelqo.”