Prompt Tracking
See how your brand shows up in AI answers, every day
Your customers now ask ChatGPT, Gemini and Google's AI answers "which X is best?". Prompt Tracking asks those questions for you every day and shows, in one panel, whether your brand appears, where it ranks, who is ahead of you and which websites feed the answers.
Measured automatically every day · Interface and reports in English, Portuguese and Turkish · Cancel any time from your panel
Measured across these AI surfaces
Visibility
62%+8
Share of voice
28%+4
Average position
2.4-0.6
Win rate
31%+5
Sentiment
78+2
Last 30 days
Share of voice ranking
- Sample Brand28%
- Competitor A24%
- Competitor B19%
- Competitor C12%
What Prompt Tracking gives you
Share of voice, position and win rate
Share among tracked brands adding up to 100 percent, position inside the answer and first-place mentions, engine by engine.
Sentiment score
Every answer that mentions you is scored 0-100: a recommendation or a mention with reservations.
Source intelligence
Sites that feed you, sites that feed your competitors, answers that cited you without naming you.
Per-engine breakdown
A separate score for every AI surface, a "why this score" prompt list, and the ads, maps and shopping cards that appeared alongside.
Full text of every answer
Kept for up to 400 days; open it on the prompt page, download it as CSV.
Alerts and weekly digest
Loss, rise, new competitor, source change; a digest e-mail every Monday.
Recommendations, opportunities, action trail
Four-part recommendations, pages where you could write, "done" marks on the chart.
Search Console and GA4
Real Google queries become prompts; visits from AI assistants, page by page.
API and MCP
Read your data in your own software or assistant; on Pro, set up and manage without opening the panel.
Continuous measurement
Why a one-off check misleads
Ask the same question today and tomorrow and the AI may list different brands. A single screenshot shows neither where you stand nor whether you are improving. Prompt Tracking measures a fixed set of questions every day, in the same market and language, and reads the results over a seven-day window: "you appeared in five of the last seven days for this question". A day that could not be measured is marked as such, never counted as zero. Every movement on the chart is a real change, not noise.
Why does AI visibility matter now?
- 87.4% of AI referral traffic comes from ChatGPT, and those visitors convert at twice the rate of traditional sources (Conductor, 2026).
- AI Overviews cut classic organic clicks by 34.5% (Ahrefs, 2025).
- Princeton research shows that adding sources, statistics and quotations to content boosts visibility in AI answers by up to 40% (arXiv, 2023).
If AI isn't recommending you, it's recommending a rival.
Last updated:
Metrics
What you measure
Visibility score
- The share of answers in your question set that mention your brand. The big number is today, next to it the average of the last seven measurements and the change arrow. Questions that already contain your name ("MessGEO pricing") are separated automatically so they cannot inflate the score, and tracked in their own section.
Share of voice, average position, win rate
- Your share among the brands you track, adding up to one hundred percent; where you appear inside the answer; how often you are named first. All three are shown per engine.
Sentiment score
- Every answer that mentions your brand gets a 0-100 score: does it recommend you or mention you with reservations? The score appears on the overview, in the Brand image tab, on every prompt and in the CSV.
Citation share
- In how many answers your site was cited as a source, your share of citations, your average source position, and the answers where your site was cited but your name never appeared.
Visibility
62%
Share of voice
28%
Win rate
31%
Sentiment
78
Sample view
Sources
Who feeds the answers? Source intelligence
AI answers do not come from nowhere; they read specific websites and cite them. Prompt Tracking keeps the full text and the source list of every answer, then answers three questions.
Your site as a source
- Your citation share, the number of answers that cite you, and your position in the source list. Your aliases and extra domains count as you.
Sites that feed your competitors
- In answers that mention a competitor but not you, which sites are cited? That list is the list of pages where your content or a comment of yours should exist.
What appeared alongside the answer
- Did Google show an AI Overview, did the answer carry ads, a map or a shopping card, did ChatGPT search the web before writing? Your mention rate is shown separately for answers with and without a search. On engines that expose it, you also see the searches the engine ran before writing.
Sites that feed your competitors
Sample view| Source | Competitor answers | You |
|---|---|---|
| guide.example | 14 answers | None |
| forum.example | 11 answers | None |
| compare.example | 9 answers | 2 |
| news.example | 6 answers | 1 |
Cited, not named
Your site was cited in 6 answers but your brand name never appeared.
Competition
Competitors and the market map
Define up to 15 competitors on Starter and 25 on Pro; visibility, share, average position and win rate side by side for each. The "competitors present, you absent" list opens the questions where you fall behind with one click. The market map groups every brand that appears in the answers: the competitors you track, candidates you have not added yet, and publishers cited as sources. When a new name rises, one click adds it to your competitor list.
Competitors
Sample view| Visibility | Win rate | |
|---|---|---|
| Sample Brand | 62% | 31% |
| Competitor A | 58% | 34% |
| Competitor B | 44% | 17% |
| Competitor C | 29% | 9% |
Action
Every measurement leads to the next step
Recommendations
- Each one has four parts: observation, evidence, likely cause, what you can try. For a finding about your site, one click takes you to the right MessGEO tool.
Opportunities
- Forum, community, list and guide pages where you could write, together with the questions they would help. Social profiles and competitor sites are left out.
Sample view
- Observation
- Competitors appear in comparison questions, you do not.
- Evidence
- 9 of 12 prompts, last 7 measurements.
- Likely cause
- Your site has no comparison content.
- What you can try
- Publish a category comparison page.
Action
Mark it done: it leaves a mark on the chart
When you apply a recommendation, mark it; the date goes onto the Score trend chart. The chart also places a numbered marker when your engine set, prompt set or tracking market changes and when the measurement method is improved. No movement is left unexplained.
Planning (Pro)
- Plan a recommendation, set a due date, get a reminder e-mail, and after you have done it see the score before and after on the same line.
Score trend
Sample view- 1New prompt set
- 2Marked done: comparison page live
Stay informed
Alerts, weekly digest, monthly report
A visibility loss, a clear rise, a new competitor in the answers, a change in the source list: four alert types, in the panel and by e-mail. A single-day swing does not trigger an alert, and an alert updates itself to "recovered" when the situation improves. Every Monday a weekly digest arrives: score change, number of questions where you were absent, the week's observations and a suggested step. On Pro, a monthly report e-mail is added at the start of the month: the last thirty days' score, citation share, your place in the share ranking and the work you marked done. You choose which e-mails you receive.
Visibility loss on 3 prompts
OngoingNew competitor in the answers
NewSource list is back
RecoveredSample view
Connections
One table with Search Console and Google Analytics
Connect your Google Search Console: the queries people really type into Google appear in the panel and become prompts with one click. Connect Google Analytics 4: visits to your site from AI assistants, page by page, next to the prompts that cite those pages. Both connections are read-only and can be cut at any time; when your plan ends the data stays, only the reading stops.
| Page | Visits from AI assistants | Prompts citing the page |
|---|---|---|
| /pricing | 214 | 9 |
| /compare | 168 | 7 |
| /blog/guide | 96 | 4 |
API and MCP
API and MCP: manage it without opening the panel
Every plan ships with the API and the MCP server. On Starter you read your panels, prompts, score series, answers, sources, competitors, opportunities and alerts in your own software or in assistants such as Claude Code and Cursor. Pro adds writing: tell your assistant "set up a panel for mybrand.com in the United States in English, add the competitors, start the first measurement"; the panel is created, prompts are added, results come back. 1,000 (Starter) or 10,000 (Pro) requests a month. Documentation in three languages.
Which prompts did I lose ground on since last week?
You lost ground on 7 prompts, mostly comparison questions, where Competitor A took your place.
curl https://api.messgeo.com/v1/panels \
-H "Authorization: Bearer mgk_..."Setup
Set up in minutes
-
1
Enter your brand and site
The wizard reads your site, works out your sector and services and finds competitor candidates.
-
2
Approve the prompt set
A set of informational, comparison and purchase questions comes ready; on the chip screen you delete, edit and add your own. The prompt report card tells you whether the mix is balanced.
-
3
Choose country and language, and let it run
Add other spellings of your name and extra domains; the first measurement starts with setup, the rest arrive every day on their own.
Prompt set
Sample view- best accounting software for small businessInformational
- invoicing app comparisonComparison
- bookkeeping software pricingPurchase
- how to choose accounting softwareInformational
- Sample Brand alternativesBrand
Local markets
Built for local markets
You choose the country and language of every panel: a company in Istanbul is measured in the Turkish context and in Turkish, a company in Lisbon in the Portuguese context and in Portuguese, a company in Austin in the United States context and in English. Interface, e-mails, weekly digest and API documentation come in English, Portuguese and Turkish. If you change the market later, the chart marks that day; the series are never mixed.
Sharing
Share, download, report
Show the panel to your manager or client with a password-protected, read-only share link; it does not count as a user and contains no answer text or settings. Every tab's CSV and a detailed daily CSV for the last 90 days are one click away; panel and prompt pages print to PDF. On Pro you download all answers with their full text in one file, and the share link can include the trend, engines, sources, alerts and actions sections.
Plans
Prompt Tracking plans
Choose a plan, connect your brand and the first measurement starts on setup.
Starter
€49/month
2 months free yearly: €490/year
- 50 prompts, 15 swaps a month
- 1 project, 1 domain
- 3 fixed AI surfaces: ChatGPT, Gemini, AI Mode
- Automatic daily measurement; your country and language
- Share of voice, average position, win rate, sentiment score
- Source intelligence: sites that feed you and your competitors
- Alerts and weekly digest e-mail
- Search Console and Google Analytics connection
- API and MCP read access; 1 share link; CSV and PDF
Pro
€99/month
2 months free yearly: €990/year
- Everything in Starter, plus:
- 150 prompts, 50 swaps a month
- 1 project, 1 domain
- Pick 3 of 6 AI surfaces
- Track up to 25 competitors
- Topic × engine heat map
- Action planning: due dates, reminders, before / after impact
- Monthly report by e-mail
- Bulk download of all answers
- 10 share links, including trend, sources and alerts sections
- Full API and MCP management: set up panels and change prompts from your assistant
Enterprise
Contact us
- ∞Unlimited projects and brands
- ∞All supported AI surfaces, unlimited model selection
- ∞Custom prompt volume and measurement cadence
- ∞Priority support and custom reporting
- ∞Onboarding and training with your team
AI surfaces are fixed on Starter; on Pro you pick 3 of the 6 surfaces yourself.
| Feature | Starter €49/month | Pro €99/month |
|---|---|---|
| Measurement scope | ||
| Tracked prompts | 50 | 150 |
| Prompt swaps a month | 15 | 50 |
| AI surfaces | 3 fixed surfaces | Choose 3 of 6 |
| Measurement frequency | Daily | Daily |
| Brand, domain, country and language per panel | 1 | 1 |
| Tracked competitors | Up to 15 | Up to 25 |
| Aliases, former names and extra domains | Up to 10 | Up to 10 |
| Setup wizard: automatic prompt set and competitor discovery | Included | Included |
| New prompts are measured right away | Included | Included |
| Metrics | ||
| Visibility score (latest and 7-measurement average) | Included | Included |
| Share of voice, average position and win rate versus competitors | Included | Included |
| Sentiment score 0-100 on every answer | Included | Included |
| Brand questions separated automatically | Included | Included |
| Per-engine breakdown and "why this score" | Included | Included |
| Visibility by topic | Included | Included |
| Topic × engine heat map | Not included | Included |
| Position distribution inside answers | Included | Included |
| Market map: every brand in the answers | Included | Included |
| Brand image | Included | Included |
| Score trend: competitor, mention and sentiment series, period picker | Included | Included |
| Chart markers: method, engines, prompt set, market, actions | Included | Included |
| Source intelligence | ||
| Your site as a source: share, coverage, position | Included | Included |
| "Cited but not named" list | Included | Included |
| Sites that feed competitors (source gap) | Included | Included |
| What appeared: AI Overview, ads, map, shopping card | Included | Included |
| The engine's background searches (where exposed) | Included | Included |
| Full text and sources of every answer (up to 400 days) | Included | Included |
| Action | ||
| Recommendations: observation, evidence, likely cause, what to try | Included | Included |
| "Done" marks, visible on the chart | Included | Included |
| Planning: due dates, reminders, before / after impact | Not included | Included |
| Opportunities: forum, community and list pages to write on | Included | Included |
| Wins and progress | Included | Included |
| Alerts and reports | ||
| Alerts for loss, rise, new competitor and source change | Included | Included |
| Alert state: ongoing / recovered | Included | Included |
| Weekly digest e-mail | Included | Included |
| Monthly report e-mail | Not included | Included |
| Notification preferences | Included | Included |
| Sharing and export | ||
| Per-tab and detailed daily CSV (last 90 days) | Included | Included |
| Bulk download of all answers (with text, last 30 days) | Not included | Included |
| PDF: panel and prompt pages | Included | Included |
| Password-protected read-only share links | 1 | 10 |
| Trend, engines, sources, alerts and actions in shares | Not included | Included |
| Connections | ||
| Google Search Console: queries and prompt suggestions | Included | Included |
| Google Analytics 4: AI traffic, page by page | Included | Included |
| API and MCP | ||
| API and MCP read access | Included | Included |
| API and MCP write access: prompts, competitors, settings, panel setup and deletion | Not included | Included |
| Requests a month (API and MCP together) | 1,000 | 10,000 |
| Requests a minute (per key) | 60 | 120 |
| Account | ||
| Interface, e-mails and docs: English, Portuguese, Turkish | Included | Included |
| 2 months free on yearly billing | Included | Included |
| Cancel from the panel; data is kept | Included | Included |
Frequently asked questions
What is prompt tracking?
Asking, every day and on your behalf, the questions your customers put to AI assistants, and recording the answers. Over time you see whether your brand appears, where it is named, which competitors are listed and which websites are cited.
Which AI surfaces are measured?
On Starter, ChatGPT, Gemini and AI Mode are fixed. Pro also opens Copilot, AI Overviews and Perplexity, and you choose 3 of the 6.
How often is it measured?
Every day, automatically. Results are read over a seven-day window; a day that could not be measured is marked as such, never counted as zero. A prompt you add is measured without waiting.
How many prompts do I need?
For most brands, 50 prompts (Starter) cover informational, comparison and purchase questions in a balanced way. With several product lines or a broad category, 150 (Pro) is more comfortable. The prompt report card tells you whether your mix is balanced, and you can swap 15 or 50 prompts a month.
Do questions that contain my brand name inflate the score?
No. Brand questions such as "MessGEO pricing" are separated automatically; the score and the share are calculated from unbranded questions only, and brand questions are tracked in their own section.
How do I track competitors?
Define up to 15 on Starter and 25 on Pro; visibility, share, average position and win rate appear side by side for each. When a new brand rises in the answers it shows up on the market map and joins your list with one click.
Which country and language is used?
You choose one country and one language per panel: Turkey in Turkish, Portugal in Portuguese, the United States in English. Measurement runs in that country context and in that language. For a second market you open a second panel.
Can I export my data?
Yes. Every tab's CSV, a detailed daily CSV for the last 90 days and PDF are in both plans; bulk download of all answers with their text is on Pro. The API gives your own software the same data.
Do I have to connect my Google accounts?
No, it is optional. Connect Search Console and real Google queries become prompt suggestions; connect Google Analytics 4 and you see visits from AI assistants page by page. Both connections are read-only and can be cut at any time.
What is the difference between Starter and Pro?
Pro gives more prompts (150) and more swaps (50), lets you choose surfaces, and adds the topic × engine heat map, action planning, the monthly report e-mail, bulk download of all answers, 10 share links and full API management. Measurement accuracy and the core metrics are the same in both.
Can I use my data with my AI assistant?
Yes. Connect assistants such as Claude Code and Cursor to the MCP server. Starter reads; Pro also sets up panels, adds prompts and changes settings from the assistant.
What happens if I cancel?
You cancel from the panel at any time. Measurement stops at the end of the period; your scores and history are kept, and when you restart you continue where you left off. Yearly billing gives you two months free.
Put your brand on the AI radar.
Sign up, connect your brand, pick your plan; the first measurement starts with setup, the rest arrive every day on their own.