Skip to main content
Quotable measures what AI engines tell your buyers. Every day it asks the engines the questions your buyers ask, in your location, records each answer and the sources behind it, and works out whether your brand was named, where, and how it was described.

The daily loop

1

You define what to track

A brand (the API calls it a project) has a name, a domain, a default location and language, up to five competitors, a set of engines, and the prompts you want tracked: the questions a buyer would type into ChatGPT or Google.
2

Quotable asks every engine every prompt

Once a day, each active prompt is sent to each engine the brand tracks, in the prompt’s location and language. Every prompt is asked as a fresh question with no earlier conversation, so answers are comparable from day to day.
3

Each answer is captured

Quotable stores the answer text, the sources the engine cited, and, where the engine exposes them, the follow-up web searches it ran to write the answer (query fan-outs).
4

Each answer is analyzed

For every answer, Quotable records whether your brand and each competitor was named in the text, whether its website was cited, the order in which brands were named, and which other brands appeared. A language-model pass then reads the answer and extracts what it says about each brand, positive, neutral or negative, and on which aspect.
5

Metrics are computed

Visibility score, mention rate, share of voice, position, citation rate and sentiment are computed from those answers. The Overview and Competitors pages can be filtered by date, engine, topic, location and prompt type, and the numbers are recalculated from the stored answers for the view you choose.

What one data point is

The unit behind every metric is one answer: one prompt, asked to one engine, on one day. A brand with 25 prompts tracked on 3 engines produces 75 answers per run. When a metric says “40% of answers”, it means 30 of those 75.

How answers are collected

Quotable collects the answer a buyer would see, not a model’s offline guess. Each answer is captured as a buyer in the prompt’s location would get it, with web search wherever the engine uses it, together with the sources the answer cites. For Google AI Overview, that is the overview shown on the Google results page for the prompt. All engines run in the prompt’s location and language. See Supported AI engines and Locations and languages.

How brands are detected

Quotable keeps two separate facts for every brand in every answer:
  • Named: the answer’s own text writes the brand’s name, as a whole word. Link addresses are stripped first, so a URL that contains your domain does not count as the engine writing your name.
  • Cited: one of the answer’s sources is on the brand’s domain or one of its subdomains.
The visibility score and mention rate count an answer when the brand is named or cited, because a link with no name still puts the brand in front of a buyer. Share of voice counts names only.
Matching ignores case. For a brand name with several words, a distinctive word that also appears in your domain counts as the brand too. For example, a brand called “Acme Analytics” on acme.com is also matched when the answer writes “Acme”. Short and generic words are never used this way.
The analysis pass extracts every brand an answer discusses, not only you and your competitors. Those discovered brands count in share of voice and in the order of mention, so your position reflects the real field of brands in the answer.

When runs happen

Each brand runs once a day, at a fixed hour between 02:00 and 07:00 in its default location’s time zone. Results usually land within the hour; some engines take longer. Adding a prompt queues a first run for it straight away, so you do not wait for the next day. See Engines and schedule. If a day’s run fails to complete, Quotable detects the gap and runs it again for that date.

Methodology you can trust

  • Same question, same conditions. Each prompt runs in the same location, language and engine every day, with no history between runs.
  • One answer counts once. A brand named five times in one answer counts as one answer for that brand, so long answers do not inflate the numbers.
  • Partial days are hidden. A date where fewer than half of the prompts produced answers is left out of trends and run-over-run changes, so a run that only covered one new prompt cannot swing your score.
  • Paid placements are separate. Ads and shopping carousels are shown in their own pages and never feed the visibility score or share of voice.
  • AI answers vary. The same prompt can get a different answer tomorrow. Read trends over several runs, use the weekly view for direction, and track enough prompts that one answer does not decide a number. See When numbers shift.

Data retention

The measurements from every answer (brand mentions, citations, positions, sentiment) and the answer text itself are kept for the life of your brand. Deleting a prompt deletes its history. Deleting a brand deletes everything tracked for it.

Next steps

Quickstart

Set up your first brand in ten minutes.

Metric definitions

Every metric, its formula and where it appears.

Glossary

The terms used across Quotable.
Last modified on November 18, 2026