From Unmeasured to #1: How Harness Institute Tracks AI Recommendations for Workshop Brands
Before June 2026 there was no number for whether ChatGPT, Perplexity, or Google AI recommended private Cursor, Claude Code, or Codex workshops. Four brands, one MCP session in Cursor, 183 tracked prompts.
What problem did Harness Institute need to solve?
Before 4 June 2026, Vasilis Tsolis and the Harness Institute team could not answer a simple question: when an engineering lead asks ChatGPT, Perplexity, or Google AI for a private Cursor, Claude Code, or Codex workshop, does our brand come up, and does the model point to our site or to a YouTube video about the vendor instead?
That is a recommendation metric, not a checkout metric, and until Reaudit there was no way to measure it. Now one operator runs four workshop-related brands from a single Reaudit MCP session in Cursor.
Why recommendation visibility is a different problem than SEO rank
Ranking #1 on Google for “Cursor workshop” and being the workshop an AI model actually recommends are not the same win. A page can rank well and still lose the recommendation to a competitor's YouTube video, a Reddit thread, or a listicle the model happens to trust more. That gap is what answer engine optimization measures, and it is why Harness Institute tracks brand mentions, source citations, and AI visibility score as three separate numbers instead of assuming a strong search position carries over automatically.
How many brands are in this case study?
| Brand | Site | Added to Reaudit |
|---|---|---|
| Cursor Workshop | cursorworkshop.com | 4 June 2026 |
| Claude Workshop | claudeworkshop.com | 5 June 2026 |
| Codex Workshop | codexworkshop.com | 5 June 2026 |
| Harness Institute | harness.institute | 6 August 2026 |
Legal AI Guide (legalai.guide), a separate documentation site also operated by Vasilis Tsolis, joined Reaudit on 11 September 2026 and runs as its own project. Its results are in a separate case study, because it is a different market and a different visibility problem: search traffic meeting AI citations, not a workshop ranking fight.
What changed after 102 days of tracking?
Before 4 Jun 2026
0
Sep 2026 (workshops)
1,003
Change
+100%
Brands onboarded to Reaudit (Jun – Aug 2026)
Cursor Workshop (4 Jun), Claude + Codex (5 Jun), Harness Institute (6 Aug).
How does Reaudit actually track this?
Each brand gets its own set of tracked prompts: the real questions a buyer would type into ChatGPT, Perplexity, or Google AI when sourcing a workshop. Reaudit runs those prompts on a schedule, records whether the brand is mentioned, extracts every citation in the answer, and scores sentiment. The Analytics Query Engine is what turns those raw answers into the ranking and citation tables in this series. The same data is available live through the MCP tools Vasilis runs from Cursor, without ever opening a separate dashboard.
183 prompts and 1,005 completed query runs is not a one-time snapshot. It is the accumulated result of scheduled runs across three brands since early June, which is also why the numbers in this series carry a publish date: they move week to week as new runs complete.
What comes next in this series?
Part 2 breaks down the numbers: #1 rankings on live commercial prompt sets, 96–99% owned-domain citation rates, and exactly who we beat on those prompts: NobleProg, DataCamp, Coursera, EY, and 46 other training competitors across the three brands.
Frequently asked questions
Who operates the Cursor, Claude, and Codex Workshop brands?
Vasilis Tsolis, through Harness Institute (harness.institute). All four brands are tracked in one Reaudit account and managed from Cursor via Reaudit MCP.
How many prompts does Harness Institute track?
Across the three workshop brands: 183 unique prompts and 1,005 completed query runs as of September 2026 (76 for Cursor, 87 for Claude, 20 for Codex).
Does Harness Institute use paid ads?
No. This case study measures AI recommendation and citation only, with no paid ad spend behind it.
What is an AI visibility score, and how is it different from citation rate?
AI visibility score is the percentage of tracked-prompt answers that mention the brand at all. Citation rate, covered in depth in Part 2, is a separate number: how often a mention also links back to the brand's own domain rather than a vendor page or a third-party video. A brand can be mentioned often and still cited rarely, which is why Reaudit reports both. See the AI visibility score glossary entry for the full definition.
What is Reaudit MCP?
Reaudit MCP exposes visibility tracking, competitor comparison, citation analytics, and content tools to AI agents in Cursor, Claude, and other harnesses. Learn more in the MCP setup guide or see real MCP use cases from other Reaudit accounts.
Related reading
Want results like Harness Institute?
See what Reaudit finds for your brand
Fifteen minutes, your own account, your real competitors, and the prompts your buyers actually type into ChatGPT, Perplexity, and Google AI. No generic deck, just your numbers.