Past
Marketers wrote pages for search engines and watched rankings and traffic.
FOUND · BE THE ANSWER WHEN THEY LOOK
F: Findable
Do we appear when relevant buyers ask?
Measured by: Share of Answer, inclusion rate, citation rate, accuracy of how AI describes the company. · Owner: your CMO
The FOUND-F agent checks, drafts, maps and records. The Strategist agent checks each draft before you see it, and nothing goes outside the company without a person approving it. See FOUND-F run on a fictional company.
PAST AND FUTURE
FOUND-F: the story in two lines
Marketers will work differently with agents. This is what changes for this letter, and what stays with a person.
Present to future
Buyers now ask an AI assistant and read one answer. The job becomes to be named in that answer, and described correctly: GEO alongside SEO.
What the agent does
Asks the agreed buyer questions through sanctioned APIs, records which companies and sources each answer names, and drafts answer-first fixes marked draft.
What the marketer or CMO decides
Which buyer questions matter, which fixes go live, and what the company says about itself.
Shifts used here: S1 From SEO to GEO; S7 From vanity metrics to share of answer. Read the full framework and sources.
How it is scored
How FOUND-F is scored
Each subcomponent is scored 0 to 5. What you tell us counts for at most 2. What our code checks counts for at most 4. Only a person who verifies the evidence can reach 5. The full method.
- InclusionMachine ceiling 2 of 5
- Answer-engine question (needs a provider key); your own answer, counted up to 2
- RecommendationNo automated check
- Your own answer, counted up to 2; a person must verify more
- AccuracyNo automated check
- Your own answer, counted up to 2; a person must verify more
- Prompt and source coverageMachine ceiling 2 of 5
- robots.txt for AI crawlers, sitemap and llms.txt
HARNESS
FOUND-F: the systems it would connect to
Asks buyer questions of answer engines through their official APIs, reads the company’s own search and analytics data, and drafts answer-first fixes for the CMS. A person publishes.
Not built yetA design, with two building blocks started. No live connector, MCP client, sign-in flow or approval queue is built, and no vendor named has approved, certified or partnered with us. The CRM read rules and pipeline story, and the Google Analytics and Search Console read rules and analytics story, run on example data only.
Systems, and how an agent would reach each
- OpenAI Responses API: read: asks the agreed questions, stores answers and cited URLs. Route: API. Checked in the vendor’s documents on 19 October 2026. (API)
- Google Gemini API with Search grounding: read: same questions, stores sources used. Route: API. Checked in the vendor’s documents on 19 October 2026. (API)
- Perplexity API: read: same questions, stores cited sources. Route: API. Checked in the vendor’s documents on 19 October 2026. (API)
- Google Search Console: read: queries and pages. Route: API, no official MCP. Checked in the vendor’s documents on 19 October 2026. (MCP)
- Google Analytics: read: landing page views and conversions. Route: MCP, read-only. Checked in the vendor’s documents on 19 October 2026. (MCP)
- WordPress, Contentful, Webflow or Sanity: write: saves a draft page only. Route: MCP. Checked in the vendor’s documents on 19 October 2026. (MCP)
Data in
- The agreed prompt set, category terms and approved claims from the strategist’s scoping record.
- Answers and citations returned by each provider API, with provider, model, mode, time, geography and prompt version.
- Search queries, pages and landing-page behaviour from Search Console and Analytics.
Data out
- Metric snapshot: Share of Answer, inclusion rate, citation rate and how accurately AI describes the company.
- Draft answer-first pages saved in the CMS as drafts, each with the buyer’s question as its heading.
- Evidence log entries and the handoff packet for FOUND-O.
Read access is the default. Writes are drafts, notes and tasks only, and a person approves each one. Spend, publishing and messages to outside people are always done by a person. How harnesses work, for all twelve.
The agent for this letter
FOUND-F does the groundwork. You decide.
Answers the question: Do we appear when relevant buyers ask? Measured by: Share of Answer, inclusion rate, citation rate, accuracy of how AI describes the company.
It works in 6 steps and stops at 3 points for a person to approve. The steps, the approvals, everything it takes in and hands back, and what it prepared in the simulation are on the agent page.
FOUND-FFindable
In
- Client profile, ICP, category terms and approved claims (from the str…
- Prompt set and category definitions agreed at scoping.
Does
- Load inputs and verify evidence from the prior cycle is intact.
- Measure the letter's metrics (Share of Answer, inclusion rate, citati…
Out
- Metric snapshot for this letter with method, provider/model, date, pr…
- Draft recommendations and drafted assets (fixes, outreach drafts, bri…
You approveStrategist approves decision-grade findings and anything published or sent externally.
What to avoid
- Do not chase every prompt, and do not build dashboards of numbers that look good but decide nothing.
- Do not copy competitor wording; differentiate instead.
- Do not scrape consumer AI interfaces or game answers.
- Do not publish without approval.
Find out how Findable looks for your company today.
Free, open and no login. Every number shows who saw the evidence.