Digital TEN X logo

Method v1.0 · 10 September 2026

How we measure AI visibility.

The scoring model behind every Digital TEN X AI Visibility Audit, published in full: eight criteria with weights that add up to 100, the test conditions, what we exclude, and what the score cannot tell you. English and Arabic are scored separately and never averaged.

  • Version 1.0
  • 8 criteria · 100 points
  • Arabic & English, scored apart

Request the free audit See the criteria

01

The model

Eight criteria. One hundred points.

Digital TEN X and our partner Agence 3A developed this assessment model together. Agence 3A uses it for businesses in Québec, Canada; this version is adapted for Qatar, and every change is listed in the changelog below. Four criteria are scored for each language; four describe the business as a whole, and both language scores share them.

AI visibility criteria, weights and what each language score reads, method v1.0
Criterion Weight English score reads Arabic score reads What it measures
01 Share of AI answers 25 English answers Arabic answers Presence rate and mention share across ChatGPT Search, Google AI Overviews and Perplexity
02 Google Business Profile 20 Shared Shared Website, phone and category; pin accuracy; rating; reviews against comparable Doha listings
03 Language coverage 15 English pages Arabic pages Key pages in the language, paired by reciprocal hreflang, read by a native speaker
04 Entity consistency 10 Shared Shared Name, Arabic registered name, phone and address agreeing across every source found
05 Answer-readiness 10 English pages Arabic pages Price, evidence, questions answered, and a direct WhatsApp or phone contact
06 Local competitive conditions 10 English answers Arabic answers 100 minus the appearance rate of the most-named rival in questions that name a place
07 Technical access 5 Shared Shared Titles, headings and indexing; AI search crawler access; structured data matching the page
08 Mobile performance 5 Shared Shared Lighthouse performance score of the homepage, mobile emulation
Total 100
02

Scoring

Two scores, never averaged.

Every criterion is scored out of 100, and higher is always better. A language score adds up each criterion score multiplied by its weight and divided by 100. The English score reads the English values of criteria 01, 03, 05 and 06, the Arabic score reads the Arabic values, and both read the same values for 02, 04, 07 and 08.

The report shows the two scores side by side. There is no combined figure: an average would let strength in one language hide a gap in the other, and in Qatar that gap is often the finding that matters.

Missing data is never scored as zero and missing weight is never shared out. If a language received no usable answers, its score is shown as incomplete, with the reason. A missing Google Business Profile is not missing data: it is a finding, and scores zero.

Language score
sum of (criterion score × weight ÷ 100)
Direction
100 is the most favourable; 0 is the most to strengthen
Rounding
Once, at the end, to a whole number. The worked example shows figures to one decimal for reading.
Work ahead
Light 67–100 · moderate 34–66 · substantial 0–33
03

Criterion by criterion

What each one measures, and how.

The tags show the points inside each criterion. Qatar signals appear only where a person can check them from public data.

  1. Share of AI answers · 25 points

    Scored separately for English and Arabic from the answers ChatGPT Search, Google AI Overviews and Perplexity give to the fixed buyer questions. Presence rate is answers naming you divided by answers received. Mention share is your mentions divided by all business mentions, counting each business once per answer. The criterion score is the mean of the two. A mention counts whether it is positive, neutral or negative; a cited source is not treated as a competitor. Your position in the answer and the sources cited are recorded, not scored.

    • Presence rate 50
    • Mention share 50
    • Per language
  2. Google Business Profile · 20 points

    Our score, not Google’s, read from the public listing. Website, phone and primary category correct: 30. Pin accuracy: 10. A pin on the building that matches the zone, street and building numbers the business publishes scores full marks, a pin in the right zone scores half, and anything else scores zero. For a service-area business, which Google asks to hide its address (S6), we check the service area instead. Public rating out of 5: 30. Review count against the median of at least three same-category Doha listings from the same search, capped at the median: 30. No listing scores zero.

    • Essentials 30
    • Pin accuracy 10
    • Rating 30
    • Reviews 30
  3. Language coverage · 15 points

    Scored per language over your key pages: the homepage, each service page you name, pricing and contact. A page earns 40 when it exists at its own address with the right language and text direction declared, 30 when it is paired with its counterpart by reciprocal hreflang (Google ignores tags that do not point both ways, S3), and 30 when a native speaker reads it as written for that language. A page the reader judges machine-translated scores zero on all three parts. The criterion score is the mean across key pages.

    • Page exists 40
    • Reciprocal hreflang 30
    • Native reading 30
  4. Entity consistency · 10 points

    The share of sources that agree on name, phone and address with the version you confirm as correct: your site copy, structured data, llms.txt, Google Business Profile, LinkedIn, and every Qatar directory or register where we find you, such as the Qatar Chamber Commercial & Industrial Directory (S7) or, for QFC-licensed firms, the QFC public register (S8). In Arabic sources the name checked is your Arabic registered name, because an English trading name and an Arabic registered name can differ, and an engine has to be able to tell they are one business.

    • Agreeing sources ÷ sources found
    • Arabic registered name
  5. Answer-readiness · 10 points

    Scored per language on up to five pages: the homepage and the pages closest to the services you name. Four signals, 25 points each: a price or price range; evidence such as a named client, a case study or reviews; questions and answers; and a contact path a buyer can use at once, meaning a WhatsApp link or a tappable phone number. The check records whether each signal exists, not how good it is.

    • Price 25
    • Evidence 25
    • Questions 25
    • WhatsApp or phone 25
  6. Local competitive conditions · 10 points

    Scored per language from the questions that name a place: Doha, Lusail, or a district such as West Bay. The score is 100 minus the percentage of those answers in which the most-named rival appears. It describes the field you compete in rather than your own work, which is why it carries less weight than your share of answers.

    • 100 − top rival %
    • Place questions only
  7. Technical access · 5 points

    Across the pages read: every page has a title and one main heading and no noindex directive, 40. robots.txt lets Googlebot, OAI-SearchBot and PerplexityBot reach the site, 40, because a page must be indexed to appear as a supporting link in Google’s AI features (S1) and a site that opts out of OAI-SearchBot is not shown in ChatGPT search answers (S4). Structured data parses and describes the business the way the visible page does, 20 (S2). Whether an llms.txt file exists is recorded, not scored.

    • Pages 40
    • Crawler access 40
    • Structured data 20
  8. Mobile performance · 5 points

    The Lighthouse performance score of your homepage under simulated mobile conditions, with the Lighthouse version and the date recorded. It is a lab test: it does not use your traffic and does not measure real visitors.

    • Lighthouse
    • Homepage
    • Lab test
04

Share of voice

The arithmetic, worked through.

Share of voice has two parts. Presence rate: answers naming you ÷ answers received. Mention share: your mentions ÷ all business mentions, with each business counted once per answer however often it is repeated.

Illustration · placeholder businesses and invented answers, not a result

Illustrative log of six answers in one language
Question Engine Businesses named, each counted once Mentions Names you
Q1 ChatGPT Search Business A, Business B, your business 3 Yes
Q1 Google AI Overviews Business A, Business C 2 No
Q1 Perplexity Business B, Business A (named twice, counted once), your business 3 Yes
Q2 ChatGPT Search Business C, Business D, Business A, Business B 4 No
Q2 Google AI Overviews No AI answer shown: excluded n/a n/a
Q2 Perplexity Business A, Business B 2 No
The calculation for the illustrative log
Step Arithmetic Result
Answers received 6 asked − 1 excluded 5
Presence rate answers naming you ÷ answers received = 2 ÷ 5 40.0%
Mention share your mentions ÷ all business mentions = 2 ÷ (3 + 2 + 3 + 4 + 2) = 2 ÷ 14 14.3%
Criterion 01 score (40.0 + 14.3) ÷ 2 27.1
Points toward this language’s score 27.1 × 25 ÷ 100 6.8 of 25
Criterion 06, if both questions name a place Business A appears in 5 of 5 answers: 100 − 100 0
05

Test conditions

How every answer is collected.

Every answer is a dated observation. Engines change, and what they say about any business can be wrong, so the conditions are fixed and the evidence is kept.

  • From Doha, no VPN

    Every question is asked over a connection in Doha, Qatar, with no VPN, proxy or remote server in between. Answers change with location, so a run from anywhere else is not comparable.

  • Private and signed out

    A private browser window, signed out, so past searches do not shape the answer. Where an engine requires an account, a new one with memory and personalisation off.

  • Fixed wording

    Each question is written once in English and once in Arabic, by a person, and asked word for word on every engine: one question per new session, no follow-ups.

  • Dated and screenshotted

    Every answer is logged with its date, time in Qatar, engine and language, captured in full, and kept unedited in its original language.

Excluded from every rate

  • Questions that name a business

    A question containing a business name tests recall, not recommendation. The question set is written without one.

  • Failed collections

    An error, a time-out or an answer that did not load is removed, not scored as zero. The report lists it.

  • No AI answer shown

    When an engine shows no generated answer for a question, there is nothing to measure. The question is excluded for that engine and listed.

06

Context

Reported, but not scored.

  • Keyword data

    Monthly search estimates for your services and areas, where they exist. Estimates are missing for some specific terms, so keywords stay context: they never block the score or change it.

  • llms.txt

    Recorded as present or absent. Google says its AI features need no AI text file (S1), and we have no primary source showing that one changes an answer, so it earns no points.

  • What public data cannot see

    Traffic, conversions, Search Console, Analytics, private profile insights, review recency and owner responses. Connecting accounts belongs to a paid engagement.

07

Evidence

What is documented, and what is guesswork.

We keep what the engines’ own documentation says apart from what the industry assumes. Source numbers refer to the list at the end of this page.

Documented by the companies themselves

  • Google Search Central · S1

    No special markup for AI features.

    Google says no additional requirements or special optimisations are needed to appear in its AI features, and no special schema.org structured data or AI text files. To be eligible as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to show in Google Search with a snippet.

  • Google Search Central · S2

    Structured data must match the page.

    Markup must be a true representation of the page content and must not describe content readers cannot see. Correct markup still does not guarantee that a search feature appears.

  • OpenAI · S4

    Opting out of OAI-SearchBot keeps a site out of ChatGPT search answers.

    OpenAI uses OAI-SearchBot to surface websites in ChatGPT’s search features. Sites that opt out are not shown in ChatGPT search answers, though they can still appear as navigational links. GPTBot, used for training, is set separately.

  • Perplexity · S5

    PerplexityBot is for search, not training.

    Perplexity describes PerplexityBot as designed to surface and link websites in its search results, and says it is not used to crawl content for AI foundation models.

Industry guesswork, and our own judgement

  • Not scored

    Ranking-factor percentages.

    Weighted lists of what makes an engine name a business circulate widely. We have found no primary source for them from ChatGPT Search, Google AI Overviews or Perplexity, so we do not score against them.

  • Not scored

    llms.txt as a signal.

    Google says its AI features need no AI text file. We record whether one exists and give it no points.

  • Scored, and labelled

    Our own weights.

    The weights on this page are our judgement of what matters in Qatar. They have not been validated as predictors of being named. The limits below say so.

08

Limits

What the score cannot tell you.

  • No predictive validation

    This is an experimental model. The weights are Digital TEN X’s adaptation and have not been tested as predictors of being named.

  • Not a forecast

    The score does not predict whether, where or when any engine will name your business. No one can promise placement in an AI answer, and we do not.

  • Our profile score, not Google’s

    The Google Business Profile sub-score is ours. It is not an official Google score.

  • Same version, same scope

    Compare two assessments only when they used the same method version and the same questions, languages and pages.

09

Version history

Changelog.

Method versions
VersionDateChange
v1.0 10 September 2026 First published version, adapted for Qatar from the source model as set out below.

v1.0 against the source model

What method v1.0 changes from the source model
AreaSource modelv1.0 for Qatar
Criteria Seven Eight
Keyword data Scored at 15; missing data blocked the overall score Unscored context; never blocks the score
Languages One requested language English and Arabic scored separately, never averaged
Share of AI answers 25; mention share 25; mean of presence rate and mention share
Google Business Profile 15; essentials 40, rating 30, reviews 30 20; essentials 30, pin accuracy 10, rating 30, reviews 30
Language coverage Not in the model New, 15
Entity consistency Not scored New, 10; includes the Arabic registered name and Qatar directories and registers
Answer-readiness 10; price, evidence, questions 10; adds a direct WhatsApp or phone contact path
Local competitive conditions 15 10
Technical access 10; title, main heading, noindex 5; adds AI search crawler access and structured data
Mobile performance 10 5

Now see your own numbers.

The free AI Visibility Audit applies method v1.0 to your business: English and Arabic, on ChatGPT Search, Google AI Overviews and Perplexity. QAR 0, no obligation.

Request the free audit The AI Visibility service