홈 / Skills / 브랜드 및 마케팅 / AI 검색 가시성 감사 AI Search Visibility Audit
브랜드 및 마케팅

AI 검색 가시성 감사 AI Search Visibility Audit

AIBarsMITSkill 언어: 영어
라이선스 확인됨스크립트 없음낮은 위험
실행 예시실제 실행 · claude-sonnet-5-5 · 2026년 10월
Use the ai-search-visibility-audit skill. I run Lumo Notes (lumonotes.example), a small offline-first note-taking app for field researchers, based in Lisbon, sold worldwide in English and Portuguese, from 9 EUR a month. Competitors: Obsidian, Notion, Joplin. AI assistants often say Lumo Notes is "a Notion plugin" and "free", which is wrong. I want to be recommended for "best offline note app for researchers". You cannot browse the web or query AI platforms here, so do not pretend to. Here is our robots.txt: User-agent: GPTBot Disallow: / User-agent: ClaudeBot Disallow: / User-agent: * Allow: / Sitemap: https://lumonotes.example/sitemap.xml The home page is a React single-page app: the HTML we serve is just <div id="root"></div> plus scripts; pricing and the feature list are rendered by JavaScript. Please start the audit.
Skill: ai-search-visibility-audit

AI search visibility audit: Lumo Notes (started 2026-10-10)

1. Scope

  • Platforms run: none. I can't browse or query AI platforms here, so there are no AI answers, no scores and no presence rate yet.
  • What I could check: the robots.txt and the home page description you pasted.
  • Target markets: worldwide, in English and Portuguese. I suggest Portugal, UK and US as the test countries. Tell me if you want others.
  • Run count: 3 runs per query per platform, each in a fresh session, logged out, with memory and personalization off.

2. Brief (what I have and what's missing)

Item Status
Product Offline-first note-taking app for field researchers. Lisbon-based. Sold in English and Portuguese from 9 EUR a month.
Competitors Obsidian, Notion, Joplin
Goal Be recommended for "best offline note app for researchers", and fix the wrong facts
Ground truth: not a Notion plugin Given
Ground truth: not free, from 9 EUR/month Given
Free trial or free tier? Missing. This matters because "free" may have come from a trial.
Other tiers, annual price, currency variants Missing
Platforms (iOS, Android, desktop, web), sync, export formats Missing
Founders or spokespeople, other owned profiles Missing
Misspellings or former names Missing
Existing third-party coverage (reviews, directories, Product Hunt, forums, Wikipedia or Wikidata) Missing

3. Site readiness findings (from what you pasted)

P1: Key facts are not in the HTML (heuristic).

  • The served HTML is <div id="root"></div> plus scripts, so price, features, location and audience exist only after JavaScript runs.
  • Vendors don't all document whether their crawlers execute JavaScript. Confirm per vendor.
  • A fetcher that doesn't run scripts would see no description at all. That leaves the AI to fill the gap from third-party pages, which could explain "a Notion plugin" and "free".
  • This is a hypothesis, not a confirmed cause. I haven't seen the answers or their sources.
  • Fix: server-render or prerender the home page, and add static pages for pricing, features, About and FAQ. These pages should state in plain sentences what Lumo Notes is, who it's for, that it works offline, that it's a standalone app and not a Notion plugin, and that it costs from 9 EUR/month. Date the pricing.
  • Owner: engineering.
  • Re-test: fetch the page with curl and check that the facts appear in the raw response.

P0 or P1: robots.txt blocks GPTBot and ClaudeBot (documented behavior, business choice).

  • Both are named in the vendors' documentation as crawlers that collect training data. Confirm that against current docs, because the names and purposes change.
  • Your User-agent: * group allows everything else. That includes OAI-SearchBot, ChatGPT-User, PerplexityBot and Google-Extended, plus any newer Anthropic user-triggered or search agents.
  • So this is a partial block, not a total one. Whether it's unintentional is your call, so I'm not deciding it for you.
  • Blocking training crawlers keeps your pages out of future model training. The cost is that models learn less from your own correct pages, so the wrong facts have fewer first-party sources to be corrected by.
  • Allowing them doesn't guarantee better answers.
  • Action: decide deliberately. If the block was intentional, keep it and make sure the search and fetch agents stay allowed.
  • Also check: whether a CDN or firewall (Cloudflare, for example) blocks these agents regardless of robots.txt. I can't see that from here.

Fine: the sitemap is declared. Check that it lists the static pages above and has current dates.

Not yet checkable:

  • Structured data (Organization and Product schema matching visible content)
  • Portuguese pages and hreflang
  • Consistency across profiles and directories
  • Freshness of old pages

Low priority: llms.txt. No vendor has documented that it affects answers.

4. Query set (fixed, 20 queries)

Brand

  1. What is Lumo Notes?
  2. Is Lumo Notes free?
  3. Lumo Notes pricing
  4. Is Lumo Notes a Notion plugin?
  5. Lumo Notes review
  6. O que é o Lumo Notes? (PT)

Category
7. Best offline note app for researchers
8. Best note-taking app for fieldwork without internet
9. Melhor app de notas offline para investigadores (PT)
10. Best note apps for academics and field researchers

Comparison
11. Lumo Notes vs Obsidian
12. Lumo Notes vs Joplin
13. Obsidian alternatives for offline research
14. Notion alternatives that work offline
15. Joplin alternatives for researchers

Problem
16. How do I take notes in the field with no signal?
17. How can I keep research notes synced after working offline?
18. Como tomar notas de campo sem internet? (PT)

Source
19. Which note apps are made in Portugal?
20. Who makes Lumo Notes?

5. Run sheet (for you to fill in)

I'll score it once you've filled it in. I won't fill it in myself.

Date Platform and mode Country and language Query # Run # Brand mentioned (Y/N) List position Accuracy vs ground truth Sentiment (quote) Competitors named Cited URLs

Platforms to cover: ChatGPT (with and without search), Claude (with and without search), Perplexity, Gemini, and Google AI Overviews.

What to paste: the full answer text for each run, especially the brand queries (1–6) and the cited URLs. The cited URLs let me trace where "Notion plugin" and "free" come from.

6. Provisional fix order (before any evidence)

  1. P1: make the facts readable without JavaScript (see above).
  2. P0, pending evidence: correct the wrong claims at their source. I can name the pages to fix once I see the cited URLs.
  3. P1: decide the GPTBot and ClaudeBot policy and check for firewall blocks.
  4. P2: add comparison pages ("Lumo Notes vs Obsidian/Joplin/Notion") and an FAQ. Keep them factual and dated.
  5. P2: look for legitimate third-party coverage, such as honest reviews, directory listings and researcher communities. Which sources matter depends on what the answers cite.

Nothing here promises that AI assistants will start recommending you.

7. Questions for you

  1. Is there a free trial or free tier, and what are the other plans?
  2. Were the "Notion plugin" and "free" claims seen on specific platforms? If so, can you paste the answers and their sources?
  3. Was blocking GPTBot and ClaudeBot intentional?
  4. Do you have profiles or listings elsewhere (Product Hunt, Capterra, AlternativeTo, GitHub, LinkedIn)?
  5. Which countries should I treat as primary?

Test run with no web access: the model cannot query AI platforms, and it said so. For a fictional app it checked the pasted robots.txt and the single-page home page, wrote a 20-query set with a run sheet and a provisional fix order, and stopped before scoring because no AI answers were supplied. It did not open any reference files, and the platform names in robots.txt should be checked against each vendor's current documentation.

하는 일

AI 답변 엔진이 브랜드를 어떻게 다루는지 측정하고 무엇을 고쳐야 하는지 알려 줍니다. 먼저 개요와 '사실 기준'(이 브랜드에 대해 무엇이 맞는지)을 받고, 다섯 그룹(브랜드, 카테고리, 비교, 문제, 출처)으로 나눈 15~25개의 질의와 플랫폼별·질의별로 최소 3회 실행하는 절차를 씁니다. 채점은 노출률, 사실 정확도, 경쟁사 대비 언급 비율, 어조, 엔진이 인용하는 출처 지도를 보고하며 하나의 만능 점수로 합치지 않습니다. 이어 틀린 정보의 출처를 추적하고, 사이트 준비 상태(AI 크롤러용 robots.txt 규칙, JavaScript로만 나오는 내용, 쉬운 답변 페이지, 개체 정보의 일관성, 구조화 데이터)를 점검해 P0~P3 수정 목록을 내놓습니다. 각 항목에는 '문서화됨', '이번 감사에서 관찰됨', '경험칙' 중 하나가 표시됩니다.

결과물

범위를 밝힌 감사: 질의 세트, 실행표, 결과표, 틀리거나 빠진 사실과 정정 방안, 출처 지도, 사이트 점검 결과, 수정 계획, 남은 질문.

이런 때 좋습니다

AI 어시스턴트가 우리 브랜드를 아는지, 올바르게 설명하는지, 추천하는지 알고 싶은 창업자, 마케터, SEO 팀. 인용이나 순위를 약속하지 않으며 AI의 답변을 지어내지도 않습니다. 플랫폼에 질의할 수 없으면 실행표를 사용자에게 넘깁니다.

참고 사항 및 위험

낮은 위험:지침만 있는 파일로 스크립트가 없습니다. 붙여 넣은 내용(robots.txt, 페이지 소스, AI 답변)만 읽고 로그인하거나 사이트를 바꾸거나 누구에게 연락하지 않습니다. 웹에 접속할 수 없는 조건에서 가상의 앱으로 한 번 시험했고, AI 플랫폼에는 질의할 수 없다고 밝힌 뒤 붙여 넣은 robots.txt와 페이지 설명을 점검하고 20개 질의와 실행표를 쓰고 채점 전에 멈췄습니다. AI 답변은 실행마다 달라서 점수는 표본일 뿐입니다. 언급되는 크롤러 이름은 각 업체의 최신 문서로 확인하세요. AIBars 자체 제작 Skill, MIT.

SKILL.md 원본패키지에 파일 2개