Top of PromptThe standard · Version 1.0 · Last reviewed
The TOP Score
A 0 to 100 measure of how often AI assistants name a business when someone asks for a recommendation in its category. This page is the whole method, including what it can't tell you.
We ask AI for a recommendation in your category, hundreds of times, across the same three assistants every time. Your TOP Score is the percentage of those answers that name you. Score 12 and you were named in roughly one answer in eight.
How is the TOP Score calculated?
Three rules keep that number honest, and they matter more than the formula itself.
- 1Only recommendation prompts count. A question like "who should I use for X in Sydney" counts. A question that names the business already, or asks for a definition, does not, because those inflate the number without telling you anything.
- 2The prompt set and engine set are held fixed between measurements. If either changes, the score is republished as a new baseline rather than compared to the old one, and we say so on the page.
- 3Every score is published with its working. The number of prompts, the number of responses collected, the engines queried and the date, so anyone can see how thin or how solid the sample is.
Which AI engines are queried?
The current engine set is the assistants Australian buyers actually use to ask for recommendations: ChatGPT, Perplexity and Google's AI answers. Each engine is scored separately as well as pooled, because they disagree with each other more than most people expect, and an average can hide a total absence on the one your buyers happen to use.
Engine coverage changes as the market does. When an engine is added or removed the version number on this page changes and the affected scores are rebaselined rather than quietly restated.
What are the three sub-scores?
The TOP Score records what appeared in our test. The diagnostic ratings highlight areas of your online presence to investigate; they do not prove why an AI system selected or omitted your business.
| Sub-score | What it measures |
|---|---|
| Readable | Can the relevant systems access your website content? Some sites are effectively invisible to AI: broken structure, missing markup, content that only appears once a human clicks something. We check whether the machines can see you at all. A diagnostic rating, not a measured output. |
| Answerable | Does your content address relevant customer questions? AI lifts answers straight off pages. If your site talks about you instead of answering what customers actually ask, there is nothing there for it to lift. A diagnostic rating, not a measured output. |
| Trusted | How is your business represented on other relevant websites? AI leans on review sites, directories and industry press to work out who is credible. We check whether the sources it trusts in your category mention you at all. A diagnostic rating, not a measured output. |
These ratings are separate assessments, not components of the measured TOP Score.
How often is it refreshed?
Businesses on an SDM program are measured monthly, because that's the cadence at which work can be reviewed and adjusted. The public TOP 100 is refreshed quarterly, and every published figure carries the date it was measured, so a number you read here is always attributable to a point in time rather than floating free.
What the TOP Score does not measure
This is the part most vendors leave out, so here it is plainly. The TOP Score is a visibility measurement, not a verdict.
- ×It is not traffic, leads or revenue. Being named in an answer is the start of a buying process, not the end of one. We connect the score to enquiries separately, through the client's own CRM, and we don't pretend the score is the outcome.
- ×It is not a quality ranking. A business can be excellent and invisible to AI, and a business can be mediocre and highly visible. The score measures what the machines say, not what the machines should say.
- ×It is not deterministic. Ask the same assistant the same question twice and you can get two different answers, which is why the score is built from many prompts rather than a single lookup.
- ×It is not necessarily what you personally see. Personalisation and location can influence signed-in results, so what you see can differ from what a stranger sees. Every measurement here is taken unauthenticated, which reduces account-specific context without removing all variation.
Why can a score move without anyone doing anything?
Because the thing being measured moves. Models get updated, retrieval behaviour changes, a competitor gets written about somewhere an assistant trusts, and any of that can shift a score without a single change to your website. That's a real limitation rather than a disclaimer, and it's why we report the trend over several quarters rather than making much of a single month.
Version and change history
| Version | Date | Change |
|---|---|---|
| 1.0 | August 2026 | First published method: formula, prompt eligibility rules, engine set, sub-scores, refresh cadence and stated limits. |
Corrections and right of reply
If you think a score we've published about your business is wrong, tell us and we'll re-run it. If the re-run disagrees with what we published, we correct the page and note the correction rather than editing it silently. The full process is on the about page.
Want your own TOP Score?
A Namecheck gives you the score, the sub-scores and the actual answers we saw, at no cost. Request one here.