Methodology · for SEO, analytics, and digital teams
How 3RD measures AI visibility, and how you can check every number.
3RD asks AI models the questions your buyers ask, run after run, in every market and language you choose. It reads every answer, scores whether you were chosen and why, and links each score back to the answers, citations, and search terms behind it.
Built by former Googlers and veterans of brands and agencies.
Development over time
Northwind visibility against the competition, all AI models
The short version
From a buyer's question to a plan, in five steps
Prompts
The questions your buyers ask AI, per topic
Runs
Each prompt sent to each model you chose, in each market, on a daily or rolling schedule
Reading
Every answer checked for brands, position, sentiment, sources, and the ads and products shown with it
Scores
Visibility, position, sentiment, citation share, rank
Diagnosis → plan
Five drivers explain the gap, and the Playbook targets it
Steps 1–4 are Monitor, and step 5 is Diagnose and Plan, the same loop the product runs.
The people behind the method
Built by people from Google and the agency world
3RD's team, with backgrounds at Google and in the agency world, designed the method, the five drivers, and the way the plan ranks work. That know-how is built into the platform, so every team gets the same reading instead of one consultant's opinion.
Peter FriisExecutive Chairman · Google, platforms, media, and transformation
Chris McCarthyCountry Manager France · More than a decade at Google in senior creative strategy across EMEA
Berna Akin ErisCountry Manager UK · Led branding business solutions at GooglePrompts
Prompts: the questions your buyers ask
3RD suggests prompts from your topics, written the way real people ask AI assistants. You can add, edit, or remove your own. Each prompt gets a tier and a target.
| Tier | What it does | Target visibility |
|---|---|---|
| Must-win | Direct commercial impact, where market share shifts | 75% |
| Core | Shapes preferences and shortlists | 50% |
| Foundational | Builds AI understanding and category clarity | 35% |
Estimated volume is a demand score from 0 to 5. It combines commercial intent, audience breadth, and question clarity, assessed in the prompt's own language.
Models, markets, cadence
Every model you choose, in every market, on a fixed schedule
Markets
A market is one country plus one language, so Switzerland–German and Switzerland–English are two separate markets, and one never hides behind the other. 3RD sends each question with its market as context.
Models
You choose which AI models answer, including ChatGPT, Gemini, Google AI Overviews, Copilot, Perplexity, Claude, and Google AI Mode. Others, such as Grok and Mistral, can be switched on for your brand.
Collection
Answers come through each AI provider's API. For ChatGPT, we can also capture answers from the consumer interface people actually use. We switch that on for you, per brand.
Cadence
Prompts run on a schedule, either daily or rolling, where each prompt takes its turn across a set window. Every run returns one answer per model, per market, and per audience, and 3RD reads every one.
Each prompt counts once per model and per market you track it in, so one question on 4 models in 1 market uses 4 prompts. See pricing
The metrics
What each number means
| Metric | Definition |
|---|---|
| Visibility | Responses that mention your brand ÷ total responses × 100 |
| Position | Your place among competitors in the answer's list, or the order of first mention. Lower is better. |
| Citation share | Citations to pages you own ÷ total citations × 100 |
| Sentiment | The confidence-weighted balance of positive and negative phrases about you, scored per response and per sentence |
| Rank | Your place when every tracked brand is sorted by visibility, e.g., "2 of 7" |
How mentions are detected: matching handles spelling variations, capitalization, possessives, and suffixes, with an AI-based fallback for ambiguous phrasing.
The five drivers
The five drivers AI uses to choose a brand
| Driver | What 3RD checks |
|---|---|
| Content coverage | Which prompts your pages answer, and where the gaps are |
| Content quality | Definitive statements, readability, key facts up front, question-and-answer patterns, and structured elements |
| Crawlability | robots.txt, rendering, schema, and speed across your site, plus an on-page and readability score per page |
| Authority | Brand mentions across the web, weighted by how often AI cites each source |
| Outreach | Third-party sources AI relies on, and which competitors they feature instead of you |
Query fan-out: the search terms AI models run on the web to answer a prompt. 3RD shows them per prompt, next to the pages AI cited.
From diagnosis to plan
How the plan is built, and how it updates
You set a goal: a metric to raise for a group of prompts, such as "Increase visibility to 70% for Employer Branding in ChatGPT." 3RD matches each task to that goal, labels it by type (content, technical, outreach, or ads), and gives it an impact estimate so the highest-leverage work sorts to the top. Technical tasks go to your SEO or web team.
Every day, 3RD automatically checks each active Playbook to see what has actually moved for the prompts in your goal. When you refresh a Playbook, 3RD reselects the tasks from the latest answers and brings back any parked task that's relevant again.
Estimated impact and measured effect are kept apart. The impact on a task is an estimate made before you act. The daily reading is what actually happened. 3RD shows both, and never presents one as the other.
Limits
What we don't claim
Answers vary.
AI models are probabilistic, and the same prompt can get a different answer tomorrow. That's why 3RD keeps asking, run after run: more answers mean clearer patterns.
Change takes time.
AI models re-crawl and re-learn on their own schedule. Expect days to weeks between publishing and a measurable shift.
Unmeasured isn't zero.
Where no answers were captured, 3RD shows a dash, not a zero, and 3RD Agent says a market hasn't run yet rather than estimating.
We only audit what you track.
Crawlability covers the domains and pages you've selected.
Attribution is partial.
Most AI answers never produce a click, so no tool can trace every AI-influenced decision. 3RD shows AI-referred traffic from GA4 next to your visibility. Pair it with a "How did you hear about us?" question for the full picture.
Check it yourself
Your data, your tools
Raw answers
Every metric opens into the actual prompts, answers and citations behind it.
Connect your analytics
Google Search Console, GA4, and Piwik PRO data sit next to your AI visibility.
Take the data out
CSV export, a REST API, an MCP server, and a Looker Studio connector.
Specialist FAQ
Questions specialists ask
Is this rank tracking for ChatGPT?
Partly. Visibility and position work like rankings across many models and markets. The difference is that 3RD also explains why, through the five drivers, and turns it into a plan.
Who designed the method?
3RD's team, including people who spent years at Google. Their know-how is built into the diagnosis and the plan, so you don't need a specialist to read the results.
How do you collect the answers?
Through each AI provider's API, with the market set as context for every question. For ChatGPT, we can also capture answers from the consumer interface. We switch that on per brand.
How many answers is a score based on?
One answer per scheduled run, per model, per market, and per audience. Two audiences in one market means two answers per model per run. Every score shows its base, e.g., "4,970 mentions in 9,780 responses."
Can we choose the competitors?
Yes. Track the competitors you choose, or the ones 3RD suggests. The competitor set defines Rank and the Leaderboard.
How long is history kept?
Your answer history stays in 3RD from the day you start tracking. We don't delete it automatically.
Start this week
Check the method on your own brand.
Seven days of full access. Pick your markets and prompts, and open any number to see the answers behind it.