Skip to content
AppraisedSI

Methodology

How domain appraisal works at Appraised SI

A domain appraisal should be explainable. Here is exactly how every report is produced — what is measured, what the AI judges, how the score is calculated, and when we will (and won't) show a number.

What an appraisal is — and isn't

An Appraised SI report evaluates the name itself: its length and structure, how easy it is to say and spell, how memorable and brandable it is, how clearly it signals a market, how suitable its extension is, and how many realistic buyers it could have.

The headline number is a quality score from 0 to 100. It is not a price. Market prices also depend on things a name-only analysis cannot observe — traffic, revenue, timing, negotiation and the specific buyer.

The rubric

Every appraisal uses the same published rubric (version 2026.1), so results are comparable across names. Each of the seven criteria is scored from 0 to 10, and the weights add up to 100%.

  • Measured criteria are computed by deterministic application code from the name (and, for extension suitability, the intended use you enter), so the same input always gets the same scores.
  • Judged criteria are scored by the AI against the anchors below. When AI is not used, they are estimated by rule-based heuristics from signals such as dictionary words, topic keywords, sound patterns and the extension, and labeled “Rule-based estimate” in the report.

  1. Length & structure

    15% weightMeasured

    How short and cleanly built the name is, measured on the label before the extension.

    Considers: Character count of the name (excluding the extension); Number of words; Hyphens and digits.

    Scoring anchors
    9–10
    Four to six characters, or a single short dictionary word; no hyphens or digits.
    7–8.5
    Seven to ten characters or a tight two-word compound.
    5–6.5
    Eleven to fourteen characters, or a longer compound that is still clean.
    3–4.5
    Fifteen to twenty characters, or three or more words.
    0–2.5
    Very long, or weakened by hyphens and mixed digits.

  2. Pronunciation & spelling

    15% weightMeasured

    Whether someone who hears the name can say it and type it correctly (the "radio test").

    Considers: Syllable structure and consonant clusters; Letter combinations with several common spellings (ph/f, c/k, x/ks); Digits, hyphens and repeated letters at word boundaries; Whether the name is made of recognizable dictionary words.

    Scoring anchors
    9–10
    One obvious pronunciation and one obvious spelling.
    7–8.5
    Easy to say; a small chance of a spelling variant.
    5–6.5
    Pronounceable but needs spelling out occasionally.
    3–4.5
    Several plausible spellings or an awkward cluster of letters.
    0–2.5
    Hard to say aloud or very likely to be mistyped.

  3. Memorability

    15% weightJudged

    How easily the name is recalled after being seen or heard once.

    Considers: Distinctiveness and imagery; Rhythm, repetition and sound patterns; Familiar words used in a fresh way.

    Scoring anchors
    9–10
    Sticks after one exposure; vivid and distinctive.
    7–8.5
    Memorable with a little repetition.
    5–6.5
    Recognizable but easy to confuse with similar names.
    3–4.5
    Generic or forgettable.
    0–2.5
    Very hard to recall.

  4. Brandability

    20% weightJudged

    How well the name could carry a credible, ownable brand.

    Considers: Distinctiveness versus descriptiveness; Tone and credibility for the intended audience; Room to grow beyond a single product; Obvious negative meanings or associations.

    Scoring anchors
    9–10
    Premium, ownable brand with broad appeal.
    7–8.5
    Strong brand candidate for its category.
    5–6.5
    Usable brand with trade-offs (generic tone or narrow scope).
    3–4.5
    Weak brand: awkward, dated or overly descriptive.
    0–2.5
    Unsuitable as a brand.

  5. Keyword & market relevance

    15% weightJudged

    How clearly the name signals a commercial topic, and how well it fits the stated use.

    Considers: Presence of commercially meaningful words; Fit with the intended industry, when provided; Clarity of what the name suggests.

    Scoring anchors
    9–10
    Exact, high-value commercial term or a perfect fit for the stated use.
    7–8.5
    Clear, relevant commercial signal.
    5–6.5
    Some relevance, or relevance depends on positioning.
    3–4.5
    Weak or unclear commercial signal.
    0–2.5
    No discernible relevance to the stated use.

  6. Extension (TLD) suitability

    10% weightMeasured

    How widely trusted the extension is, and how well it matches the intended use.

    Considers: General recognition of the extension; Topical fit (for example .ai for AI products); Geographic targeting of country-code extensions; Registration restrictions.

    Scoring anchors
    9–10
    .com, or a topical extension that is an exact fit for the stated use.
    7–8.5
    Widely recognized extension with good topical fit.
    5–6.5
    Recognized extension with a narrower or regional audience.
    3–4.5
    Less familiar extension; limited buyer pool.
    0–2.5
    Obscure or restricted extension for this use.

  7. Commercial breadth

    10% weightJudged

    How many realistic businesses or projects could use the name, and how large that buyer pool is.

    Considers: Number of plausible industries and use cases; Size and spending power of likely buyers; Whether the name is tied to a fad or a narrow niche.

    Scoring anchors
    9–10
    Many well-funded potential users across industries.
    7–8.5
    Several strong use cases.
    5–6.5
    A handful of realistic use cases.
    3–4.5
    Niche appeal.
    0–2.5
    Very few realistic users.

How the total is computed

The weighted total is calculated by Appraised SI's code — never by the AI.

total = round(Σ(score × weight) ÷ 10)

Scores are clamped to 0–10 and rounded to the nearest half point first. Weights are percentages, so a name that scores 10 on every criterion totals 100.

Worked example

For an illustrative set of scores (not a real domain), each criterion contributes score × weight ÷ 10 points:

Worked example of the weighted total
CriterionScoreWeightPoints
Length & structure815%12
Pronunciation & spelling915%13.5
Memorability715%10.5
Brandability7.520%15
Keyword & market relevance615%9
Extension (TLD) suitability1010%10
Commercial breadth6.510%6.5
Sum 76.5, rounded77/100

A total of 77 falls in the Strong band. Grades map to totals like this:

Grade bands
GradeTotalMeaning
Exceptional85–100Rare combination of brevity, clarity and brand strength.
Strong70–84A strong name with few meaningful weaknesses.
Solid55–69A usable name with clear strengths and some trade-offs.
Moderate40–54Workable, but with notable weaknesses.
Limited0–39Significant weaknesses for most uses.

The role of AI

When an AI model is configured, it receives the domain name, the measured facts and measured scores, your chosen perspective and — only if you provide them — your intended use and notes. It is instructed to score the four judged criteria against the published anchors, write the narrative sections, and return structured data.

  • It is told never to state or guess traffic, backlinks, search volume or rankings, revenue, sales history, registration or availability status, domain age or trademark clearance; never to describe anything as verified or guaranteed; and never to include links or contact details.
  • Its answer is validated against a strict schema: scores are clamped and lists are capped. If the format is wrong, the model gets one chance to correct it.
  • A best-effort filter then removes sentences that still appear to make those claims, or that contain links, contact details or prices outside a labeled AI-estimated range. It catches common phrasings, not every possible one, so read the narrative as the model's opinion rather than as verified fact.
  • If the AI is unavailable, we tell you. We never switch to a different provider silently. You can choose a rule-based analysis instead, and the report says so.
AI is configured on this site

Reports labeled “Live AI analysis” include AI-judged criteria. Rule-based quick scores (used in bulk quick mode or when you choose a rule-based analysis) are labeled “Rule-based quick score”.

Value outlook and numbers

Every report includes a value outlook: a qualitative level computed in code from the total score and, for the top level, the extension. It is a judgment about demand, not a price.

High
A total of 82 or more on a classic generic extension (such as .com, .net or .org) or one popular with technology brands (such as .ai, .io or .co).
Promising
A total of 68 or more, including names that would reach High on a more familiar extension.
Moderate
A total of 52 to 67.
Limited
A total below 52.

Criteria scoring 7.5 or more are listed as the main drivers, and criteria scoring 4.5 or less as detractors (up to four of each).

A report can contain three different kinds of numbers. Each has its own label and look:

  • Owner asking priceShown only when the name is listed in the Appraised SI marketplace. It is set by the seller and is never part of the appraisal.
  • Verified-sales estimateA range computed in code from the owner's dataset of verified comparable sales, each with a price, currency, date and public source.
  • AI-estimated rangeThe AI model's opinion based on the name's characteristics, labeled as a model opinion — not a market valuation.

When a verified-sales estimate is shown

  • Only sales in US dollars from the last 15 years are considered.
  • A sale is relevant when it has the same extension and either the same structure (for example one word, or letters and digits) with a length within three characters, or a shared dictionary word of at least four letters.
  • At least three relevant sales are required. The range is the middle half (25th to 75th percentile) of their prices, rounded to two significant figures, and up to 10 of the sales used are listed with their source.
  • No verified comparable-sales data is configured on this site yet, so this estimate is not shown.

When an AI-estimated range is shown

It is enabled on this site and appears only on reports produced by a live AI call — never on rule-based or demo reports.

When there is not enough evidence for a number, the report says so plainly. We would rather show no number than an unsupported one.

What we don't check

  • Traffic, type-in visitors, revenue or existing website history.
  • Backlinks, search-engine history or keyword search volume.
  • Trademark conflicts — check official trademark databases and consider professional advice.
  • Registration status and domain age.
  • Whether a name is available to register or buy. An appraisal is not an availability check.

Limits, caching and privacy

  • To keep the service fast and fair without accounts, each visitor can make up to 60 appraisal requests per 10 minutes. Live AI appraisals are further limited to 6 per 10 minutes and 30 per day for each visitor, within an overall daily capacity for the site. Daily limits reset at midnight UTC, and results served from the cache don't count toward them.
  • Identical requests (same name, perspective, intended use and notes) are served from a cache for up to 168 hours, so repeat lookups are instant and consistent.
  • Reports are stored for 30 days so they can be shared by link, then deleted. At most 100,000 are kept; beyond that, the oldest are deleted first.
  • Bulk appraisal accepts up to 10 domains per run.

Read the privacy notice for what is stored and for how long.

Informational only

Appraisals are informational estimates. They are not financial, investment, legal or trademark advice, and they do not guarantee any sale price.

Appraise a domainCompare domains