ai-search-strategy
Skill1 starsChanged 3 days agoMITView on GitHub
Own the AI search strategy for a business: read where AI apps mention it and where they mention rivals, fix what is broken, and run a small set of repeated systems (listing outreach, answer pages, search-term coverage, head-to-head upkeep). Use for "what should I do about AI search", "run my AEO this week", a weekly AI search review, or a first plan for a new project. Start here when no other skill fits. Works from a count by hand, or from a Frontmost AI project.
In this repository18 other files it gives its agents.
- ai-readinessSkill
- answer-accuracySkill
- answer-pagesSkill
- community-repliesSkill
- customer-surveySkill
- directory-listingsSkill
- listicle-outreachSkill
- mention-checkSkill
- name-checkSkill
- page-draftingSkill
- placement-trackerSkill
- question-draftingSkill
- question-gapsSkill
- rival-gapsSkill
- scan-reviewSkill
- site-auditSkill
- source-mapSkill
- tactic-checkSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
16 rules. About 0 are common, 9 are distinctive, and 7 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Use this guidance to decide what a business does about AI search this week, and to keep the work going week after week.
- Distinctive: line 14
With a Frontmost AI project: do every step.
- Distinctive: line 15
With no project: use the skills that need no account, in this order:
- Very specific: line 22
Read "AI Visibility" as the mention-check score.
- Very specific: line 29
If a tool returns setup_required, give the person the url from the result.
- Very specific: line 33
Call list_projects, then get_scan for the latest scan.
- Very specific: line 35
Call get_rivals. Record the rivals that AI apps name most.
- Very specific: line 36
Call get_sources with questions: "unbranded".
- Very specific: line 39
Check that the questions are the right ones before you trust the numbers.
- Very specific: line 70
Use the scan-review skill for what changed since the last scan.
- Distinctive: line 72
Read movement over a window of two to four weeks.
- Distinctive: line 73
Stop a system when its evidence is gone.
- Distinctive: line 78
Never say that a system works, and never promise a result.
- Distinctive: line 79
Keep done, sent, published, and mentioned as separate facts.
- Distinctive: line 80
Do not send, publish, or spend without the person's approval.
- Distinctive: line 82
If the data is thin, such as a first scan, say so.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AI search strategy
Use this guidance to decide what a business does about AI search this week, and to keep the work going week after week.
One action rarely changes what AI apps say. Repeated work does. So this skill fixes what is broken first. Then it picks systems: small jobs that repeat on a rhythm, each fed by a part of the data.
With or without Frontmost AI
- With a Frontmost AI project: do every step.
With no project: use the skills that need no account, in this order:
site-audit, for what is broken on the site.question-drafting, for the buyer questions.mention-check, for a count of AI answers by hand.source-map, for the pages that shape those answers.
Then pick systems in step 3 from that count:
- Read "AI Visibility" as the
mention-checkscore. - Listing outreach: the evidence is a
source-maprow with the result Not present. - Answer pages: the evidence is a question group with no Mentioned result. Use
page-drafting. - Skip search-term coverage and head-to-head upkeep. Both need recorded answers.
Say that the count is a small sample, and pick one system at most until a second count agrees. Step 4 uses the next
mention-checkin place of a scan.
If a tool returns setup_required, give the person the url from the result. If a tool returns plan_required, say that this data needs a paid Frontmost AI plan. In both cases, continue with the no-project path.
Step 1: Read the data
- Call
list_projects, thenget_scanfor the latest scan. Passquestions: "unbranded"to count only the questions that do not name the business. - Record AI Visibility with its counts: the answers that name the business, of the answers collected. Record each AI app separately.
- Call
get_rivals. Record the rivals that AI apps name most. - Call
get_sourceswithquestions: "unbranded". Record the pages that answers cite, and whether each page lists the business. - Call
get_readiness.
Check that the questions are the right ones before you trust the numbers. If most questions name the business, or few match how buyers ask, use the question-drafting skill first.
Step 2: Fix what is broken
A broken thing is a single fix. It is not a system. Do these before anything else:
- A blocked crawler or a page that AI apps cannot read. Use the
ai-readinessskill for a fast check, or thesite-auditskill for the complete site. - A wrong fact about the business in an AI answer: a wrong price, a wrong product, or another company with a similar name. Use the
answer-accuracyskill to list the answers with a fact to check. The person marks what is wrong. The correction is a clear statement of the fact on the business's own page. For another company with a similar name, use thename-checkskill.
Step 3: Pick the systems
Pick a system only where the data shows its evidence. Most businesses run one or two.
| System | Evidence in the data | The repeated job | Rhythm | Skill |
| Listing outreach | A cited page lists rivals and not the business | Pitch the most cited such page | Each week | listicle-outreach, placement-tracker, source-map |
| Answer pages | A question or topic where answers name rivals and not the business | Publish or extend one page that answers it | Every two weeks | question-gaps, answer-pages, page-drafting |
| Search-term coverage | A web search that ChatGPT ran (get_queries), where the answers do not name the business | Add that subject to a page that exists | Each month | answer-pages |
| Head-to-head upkeep | A claim in a comparison answer that the site can rebut, or a rival that moved | Answer the claim on the comparison page | After each rival move | rival-gaps |
For each system that you pick, state:
- The evidence, with counts and the scan date.
- The first job, exact enough to do today: the page, the question, or the claim.
- Where the next job comes from. Each system has a feed: Sources, Questions, Queries, or Rivals.
- The measure to read later, and when.
Reddit threads are a separate job. Use the community-replies skill. Directories and business profiles are a separate job too. Use the directory-listings skill.
Step 4: Review each week
- Use the
scan-reviewskill for what changed since the last scan. - For each running system, report the jobs done, the jobs open, and the measure. For listing outreach, the measure is live listings, then the later answers that cite those pages.
- Read movement over a window of two to four weeks. The same question does not return the same list of businesses twice, so one scan is a first read.
- Stop a system when its evidence is gone. Start one when new evidence appears.
Rules
- Give every number with its counts and its date.
- Never say that a system works, and never promise a result. Say what the data supports, and what you will read to find out.
- Keep done, sent, published, and mentioned as separate facts.
- Do not send, publish, or spend without the person's approval. Each skill states what needs it.
- This skill sets no schedule. To repeat the review, the person runs a prompt on their own schedule.
- If the data is thin, such as a first scan, say so. Give the fixes and one system at most.
Output
- Where the business stands: AI Visibility with counts, for each AI app, and the two or three rivals that answers name most.
- Fixes, in order, or "nothing is broken".
- The systems to run, each with its four points from step 3.
- The first job to do today.
- What you could not check.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

