Track · Daily Prompt Tracking
Daily AI prompt tracking with three levels of proof.
Spyglasses runs your tracked prompts every day against ChatGPT, Gemini, and Google AI Overviews, with Claude and Perplexity available as an add-on. What comes back works at three levels: top-line KPIs you can share, backing metrics you can defend, and raw per-citation data you can take anywhere, in the app or through the MCP server.
Daily prompt tracking at a glance
- Prompts run every day across ChatGPT, Gemini, and Google AI Overviews; Claude and Perplexity are available as an add-on.
- Three metric tiers: shareable KPIs, backing metrics, and raw per-citation data.
- Goals match citations automatically, including retroactively when you add one later.
- A white-labeled summary email lands every Monday; raw data is available through MCP.
Share of goal influence
42.3%5.1%increased, improved
- Gear review roundups24 citations · 18.5%
- /guides/* page group19 citations · 14.6%
- Outdoor trade press12 citations · 9.2%
- Jul 20 · ChatGPT · “best 4-season tents for alpine conditions” → outdoorgearlab.com/tent-roundup
- Jul 20 · Perplexity · “which tent brands have the best warranty” → acmeoutdoor.example/guides/warranty
- Jul 19 · Gemini · “ultralight backpacking setup for beginners” → switchbacktravel.example/ultralight
A series, not a snapshot
One-off checks mislead. A daily series doesn't.
AI answers change between sessions and days. A single check is a coin flip; a daily series is a trend. Spyglasses runs every tracked prompt on a fixed daily schedule so your numbers come with enough samples to mean something.
How many prompts make a reliable number? Try the calculator.
Reliability calculator →Tier 1 · Share it
KPIs you can put in front of a board
Share of goal influence and attributed brand mentions, with period deltas. A white-labeled summary email goes out every Monday, and exports ship as web pages, PDF, or PPTX in your brand.
Tier 2 · Defend it
Every number opens to show its work
Click any KPI and the backing metrics are right there: influence rates per goal, weekly trends, and the citations behind each count. When a client asks “says who?”, you show them.
Tier 3 · Take it anywhere
Raw data, in the app or over MCP
Every citation row carries its prompt, platform, date, and answer context. Analysts pull the same rows through the MCP server into notebooks, BI tools, or custom client workflows.
Analysts: the whole account is queryable from Claude or any MCP client.
The MCP server →Goals
Track what you're trying to move
A goal names the outcome: your pages, a publisher, an article, an author, or a whole coverage group. Every matching citation counts toward it, including past ones when you add a goal later.
- Page group/guides/*19 · 14.6%
- PublisherOutdoorGearLab24 · 18.5%
- Article“Best tents of 2026” roundup9 · 6.9%
- AuthorRiley Chen (trade press)5 · 3.8%
- Coverage groupSpring launch coverage (32 URLs)12 · 9.2%
Personas
Same question, different person
AI doesn't give everyone the same answer. A Medicare senior in Syracuse and a parent managing three prescriptions get different pharmacy answers. A homeowner in Minneapolis after a hailstorm hears about different roofers than a landlord does. A CFO and an IT lead get different software shortlists. Spyglasses drafts personas from your brand profile, so they match the people who ask about you. Add them to a project and each discovery prompt also runs every day as each persona, on ChatGPT and Gemini. Perspectives then shows where AI makes you its top choice, where it cautions someone against you, and who it recommends instead.
Persona answers sit next to your headline numbers and never change them.
How personas work →Works with everything you track
People, places, and products too
Start tracking today.
Set up prompts now; the first daily run lands by tomorrow morning.
Daily prompt tracking FAQ
Common questions about how daily tracking works.
When do prompts run?
Every day, on a fixed schedule, on ChatGPT, Gemini, and Google AI Overviews, plus Perplexity and Claude if you add them. Runs happen overnight, US time, and results are in your dashboard by morning.
How many prompts do I need for reliable numbers?
It depends on the margin of error you can accept. More prompts and more nights mean tighter numbers; our reliability calculator shows the math for your sample size.
Can clients see the data behind a KPI?
Yes. Every KPI drills down to per-goal influence rates and the individual citations behind each count, with the full AI answer a click away.
Can I export raw prompt and citation data?
Yes. In the app you can drill to raw rows, and the MCP server exposes the same data to Claude or any MCP client for exports, notebooks, and BI tools.
What's a goal and how does it get hit?
A goal names what you want cited: a page group, publisher, article, author, or coverage group. Any tracked citation that matches counts as a hit, and new goals match past citations retroactively.
Do unused prompts cost me anything?
No. Only prompts assigned to an active project run each day and count toward your plan's usage.
Can I see how AI answers different kinds of customers?
Yes. Personas are drafted from your brand profile and tracked prompts, so they match who asks about you: a mix of people, like a Medicare senior or a homeowner after a hailstorm, or a team at one organization, like a CFO and an IT lead weighing the same purchase. Add them to a project and each discovery prompt also runs once a day as each persona on ChatGPT and Gemini. Each persona run counts as one tracked prompt, like a location. Your headline metrics count only the regular answers, so they never move.