How to track your AI visibility without a subscription
You can track AI visibility on your own. List twenty buyer questions, ask each assistant on a schedule, and log who is named and linked. Segment assistant referrers in analytics. Read the pattern and fix your pages that assistants can cite.
By Théophile Louvart, founder of Porteur · Updated 14 September 2026 · Markdown
The method in five minutes
- Write twenty buyer questions for your product. Example: “best uptime monitoring for startups”, “yourproduct vs rivalname”, “pricing for log management”.
- Ask the same list in ChatGPT search, Perplexity, Google AI Mode, Claude, Gemini and Microsoft Copilot.
- Log three things for each question: who is named, who gets a clickable link, and the order. Paste the cited URLs.
- Repeat weekly. Compare week to week. Look for named, named without a link, or absent while a rival is named.
- Add a referrer segment in analytics for chatgpt.com, perplexity.ai, copilot.microsoft.com and gemini.google.com. Watch sessions and signups from those.
- Keep your robots.txt open to the answering and search crawlers listed below. If they cannot fetch you, you cannot be cited.
Pick twenty buyer questions that assistants already answer
Assistants cite pages that answer a buyer’s question plainly. Start with specific, high-intent questions you want to win. Cover categories, comparisons and problems.
- Category picks: “best [product type] for [use case]”, “top [product type] with [capability]”.
- Comparison: “[yourproduct] vs [rival]”, “alternatives to [rival]”, “cheaper than [rival]”.
- Jobs to be done: “how to [do task] with [product type]”, “secure [thing] on a budget”.
- Buying details: “[product type] pricing”, “is [product type] GDPR compliant”, “self hosting [product type]”.
- Proof: “who uses [product type]”, “case study [use case]”.
- Branded: “[yourproduct] review”, “[yourproduct] documentation [topic]”.
Aim for twenty. If you are thin on ideas, read People Also Ask on Google and the headings on comparison pages that already rank. Use the same wording you see buyers using.
Set up the sheet and schedule
Make a tab per assistant
Tabs: ChatGPT, Perplexity, Google AI Mode, Claude, Gemini, Copilot. Columns: Date, Question, Assistant answer title if any, Named brands in order, Links present Y/N, Cited URLs, Your rank among named, Notes.
Add a summary tab
One row per question. Columns: Named with link count, Named without link count, Absent count, Rivals most cited, Pages to fix, Status.
Set cadence
Weekly is enough for a small site. Put a thirty-minute slot in your calendar. If you ship a launch or a big page, run an extra cycle after indexing.
Define a naming rule
Use your product name exactly, as on your site. Assistants are literal. “Acme Monitor” and “Acme Monitoring” read as different entities.
A filled row for “/pricing” might read: named without link in ChatGPT and Perplexity, absent in Claude, named with link in Copilot. That points to gaps you can close.
Ask each assistant the same way, then log names and links
Generative engine optimisation means being named and cited in answers from ChatGPT search, Perplexity, Google AI Overviews and AI Mode, Claude, Gemini and Copilot. There is no submission form. They pull from web indexes and cite pages that answer cleanly.
- ChatGPT search: type your question. If it does not search, click Search. Copy the Sources panel and inline citations. The crawler does not execute JavaScript, so log server-rendered URLs.
- Perplexity: ask once in normal mode, once with Pro Search for key questions. Copy the numbered citations and the sources list.
- Google AI Mode: open the AI tab after a search. Copy the linked sources in the prose answer. If there is an AI Overview instead, use that.
- Claude: ask and copy inline citations. If it opens web results, log those first.
- Gemini: ask and check for links beneath the answer. If it searches, it grounds with Google Search and shows links. Copy them.
- Microsoft Copilot: ask in the browser or app. Copy the cited links drawn from Bing’s index.
Keep the crawlers that power answers allowed
If the answering and search crawlers cannot fetch your pages, you will not be cited. robots.txt is honoured voluntarily, and blocking training does not remove what is already collected. Keep answer crawlers in, even if you block model training.
| Vendor or assistant | User agent token | What it is for as of 2026 | Block effect |
|---|---|---|---|
| OpenAI ChatGPT search | OAI-SearchBot | Indexing for ChatGPT search results | Blocks appearance in ChatGPT search |
| OpenAI on-demand fetch | ChatGPT-User | Fetches a page when a user asks | May prevent fetching your page during an answer |
| OpenAI training | GPTBot | Crawling for model training | Does not affect search citations |
| Anthropic Claude search | Claude-SearchBot | Search index for Claude | Blocks appearance in Claude search answers |
| Anthropic on-demand | Claude-User | Fetch on user request | May prevent fetching your page during an answer |
| Google Search and AI | Googlebot | Search crawling, sources for AI Overviews and AI Mode | Blocks Search and AI citations |
| Google Gemini control | Google-Extended | Training and grounding control for Gemini apps | Does not affect Search or AI Overviews |
| Perplexity index | PerplexityBot | Builds Perplexity’s index | Blocks appearance in Perplexity answers |
| Perplexity on-demand | Perplexity-User | Fetches pages to quote | May prevent fetching your page during an answer |
| Microsoft Bing and Copilot | Bingbot | Bing index and Copilot sources | Blocks Bing and Copilot citations |
| Apple | Applebot | Siri and Spotlight | Blocks Apple surfaces |
| Apple training control | Applebot-Extended | Apple Intelligence training control | Does not affect Siri or Spotlight |
| Common Crawl | CCBot | Open crawl used to train models | Blocks new training data |
| Meta | meta-externalagent / meta-externalfetcher | Training and fetcher | Blocks new collection |
| Amazon | Amazonbot | Amazon crawler | Blocks Amazon’s crawling |
| ByteDance | Bytespider | ByteDance crawler | Blocks ByteDance crawling |
| Cohere | cohere-ai | Cohere crawler | Blocks Cohere crawling |
| DuckDuckGo | DuckAssistBot | DuckDuckGo assistant | Blocks Duck Assist fetching |
Allow OAI-SearchBot while disallowing GPTBot if you want citations in ChatGPT search but no training. Allow Claude-SearchBot, PerplexityBot, Googlebot and Bingbot to be visible in assistants.
Segment assistant traffic in analytics
Visits from assistants arrive with referrers you can segment. This is your bottom-line check that citations send traffic. Google’s AI surfaces arrive as google.com, mixed with other search clicks.
- Create a segment or report for referrers: chatgpt.com, perplexity.ai, copilot.microsoft.com, gemini.google.com. Add bing.com and google.com for context.
- Track assisted signups and trials from those referrers. Add UTM tags on your own Pages or Spaces on Perplexity if you use them.
- Watch changes after a content update or a new comparison page. Does ChatGPT or Perplexity send you more sessions within 28 days of shipping?
- Compare the referrer trend to your sheet. A rise in “named with link” should precede a rise in assistant referrers. If not, the link may be low in the sources list.
Search Console reports clicks from AI Overviews and AI Mode inside normal web clicks. There is no separate breakdown as of 2026. Track those queries like any other in the Performance report.
What tracking products automate and when they are worth it
Manual tracking gets you signal within an hour a week. Tracking products automate the asking, logging and change alerts. They are worth it when your list of questions, markets or rivals outgrows a sheet, or when a team needs alerts rather than a calendar slot.
- They ask your saved questions on a schedule across assistants and regions.
- They log named entities and cited URLs automatically.
- They alert you when you gain or lose a citation or when a rival displaces you.
- They add share of voice and time-series charts you would otherwise build by hand.
- Some connect to Search Console and analytics to join citations to clicks and conversions when you grant access.
Use a product when you run more than twenty questions across more than two markets, or when you manage several brands. Stay manual if you have one product and one market and you enjoy seeing the raw answers weekly.
Read the result: named, named without a link, absent
You will see three states. Treat each differently, and fix one thing at a time. The aim is to be named with a link for your buying questions in as many assistants as possible.
- Named with link: protect and expand. Check the cited URL. Ship a cleaner version of that page with the answer in the first paragraph, clear headings that match the question and a visible date and author.
- Named without link: earn the citation. Assistants often cite lists and comparisons. Get your product onto the third-party list they already cite, or ship your own structured comparison that answers the question plainly.
- Absent while a rival is named: change the source set. Find which pages the assistant cites. Pitch those editors for inclusion, or build a page on your site that mirrors the structure and scope they favour and interlink it from /pricing, /features and relevant guides.
A fixed page looks like “/guides/best-uptime-monitors-for-startups”: H1 that matches the query, a summary table in the first screen, consistent product names, a date, and links out to rivals and sources the assistants already cite.
Make your site easy to cite
Search-backed assistants retrieve from web indexes and reward structure and clarity. Write for quotation. Add the pages that assistants and the lists they cite already favour. Keep JavaScript from hiding content.
- State the answer in the first two paragraphs. Example on “/pricing”: first lines list plans, who each is for and the monthly price.
- Use headings that match questions: “Is it SOC 2 compliant”, “Does it have an API”, “Can I self host”.
- Add a summary table for comparisons. Include features, limits, and who each product is for. Use exact product names.
- Use consistent product and company names across your site, docs and pricing. Do not alternate “Acme Monitor” and “Acme Monitoring”.
- Add a visible date and an author. Update the date when you refresh numbers or screenshots.
- Link out to sources assistants already cite. If they quote a study or a standard, cite it the same way.
- Ensure core content renders server side. ChatGPT’s crawler does not execute JavaScript. If your answer appears only after scripts run, it will not be read.
- Get listed on roundups that already rank. If AI answers cite “Top X tools” posts, ask for inclusion or provide data that merits it.
Cadence, thresholds and ownership
Give AI visibility a small, regular slot. Aim to move one question from absent to named each month. Fold fixes into your normal shipping rhythm. Keep ownership simple.
- Weekly: run the sheet, scan referrers, file one fix ticket. Example: add a summary table to “/alternatives/rivalname”.
- Monthly: review the summary tab. Pick two questions to push from named without link to named with link. Plan two pitches to cited roundups.
- Quarterly: audit robots.txt for the crawler tokens above, and ship or refresh two comparison pages or buying guides.
- Assign one owner. Add a deputy for holidays. Keep the questions list in version control next to your site content.
Questions
Write twenty buyer questions, ask them in ChatGPT search, Perplexity, Google AI Mode, Claude, Gemini and Copilot every week, and log who is named and linked. Segment assistant referrers in analytics to see traffic. Keep a summary of named with link, named without link and absent.
They automate asking your saved questions across assistants, log names and links and alert you to changes. Use one when you track many questions, several markets or multiple brands, or when you need alerts instead of a weekly slot. Stay manual if you have one product and one market.
Review robots.txt. Allow OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot and Bingbot. You can disallow GPTBot and still appear in ChatGPT search by allowing OAI-SearchBot. Blocking answering and search crawlers makes you invisible to assistants.
Clicks arrive inside normal web search clicks in the Performance report. As of 2026 there is no separate breakdown. Track the affected queries and pages like any other search click and compare to your sheet of citations.
Treat it as progress. Find which sources are cited in that answer and get your product on those lists and comparisons. Ship or improve a page on your site that answers the same question in the first screen with matching headings and a summary table.
Assistants sit above results for many buyer questions. If a rival is named or linked and you are not, they will earn the click and the trial. Tracking shows you which pages to fix and which lists to join.
Sources
Check my site, free
Paste your URL to get a free AI visibility check that reads your site, the searches around it and the rivals on them in about thirty seconds and shows three findings whole. Porteur can connect to Search Console and analytics later if you choose.
- Free check, no card
- Read-only, your own accounts
- Readable by your agent
Read next
- GuideGenerative engine optimization (GEO): how to be named by AI answers
- GuideHow ChatGPT search picks and cites sources
- GuideHow Perplexity picks its sources, and how to become one
- GuideGoogle AI Mode: what it is and how a small site gets cited
- Guiderobots.txt for AI crawlers: GPTBot, ClaudeBot, PerplexityBot and what to allow
- GuideWriting pages AI assistants can cite
- GlossaryShare of voice
- GlossaryBrand mention