I launched a new product two months ago. How do I know when AI assistants start recommending it?

Use two kinds of evidence: whether assistants name your product, and whether that attention brings visitors or customers. Neither alone tells the whole story.

1. Make a list of real buyer questions. Include prompts like “What’s a good [product category] for [use case]?” and “Compare [your product] with [competitor].” Use the wording your customers actually use, not just your brand name.
2. Check those prompts regularly in the assistants your buyers use. Record whether your product is recommended, where it appears, which competitors are named, and any links or sources shown. Keep the prompts and conditions consistent; answers can vary, so a single result isn’t proof of a lasting change.
3. Use an AI-visibility tracker if you want this automated. Tools such as Semrush offer prompt tracking across selected AI platforms. Treat the results as a sample of answers to monitored prompts—not a complete count of every recommendation assistants make. (semrush.com)
4. Track the visits and outcomes you can observe. In your analytics, look for referrals from AI assistants and measure sign-ups, purchases, or other conversions from those visits. OpenAI says ChatGPT search referral URLs include utm_source=chatgpt.com, which can help identify some traffic. But referral analytics only catch people who click through; they won’t reveal every recommendation, and some visits may be unattributed. (help.openai.com)
5. Ask new customers how they found you. Add “Did an AI assistant recommend us?” to a signup or checkout survey. It can catch influence that web analytics misses.

Since your product launched two months ago, start a baseline now and compare week to week. Track recommendation rate, competitor mentions, citations, AI-referred visits, and conversions. The strongest signal is not just “an assistant mentioned us,” but “it keeps recommending us for relevant questions, and that visibility is associated with qualified customers.”

There's no alert from AI platforms when they start mentioning a product, so you have to check for it yourself. The usual approach is to run a fixed set of questions on a schedule and log the answers.

1. Set up a manual test (free)

• Write 15 to 25 questions a real buyer would ask, such as "best [category] tool for [use case]," plus a few branded ones like "is [product] good for [use case]?" Unbranded prompts show whether you're discovered at all. Branded prompts show how the AI describes you.
• Run the same list on ChatGPT, Gemini, Perplexity, Claude, and Google AI Overviews. One guide recommends running branded and unbranded test prompts across multiple LLMs weekly.
• Log whether you're mentioned, your position in the answer, which competitors appear, and which sources are cited.
• Answers vary from run to run, so a single screenshot can mislead. One analysis notes that every AI visibility number is a sample from a changing response distribution, not a permanent rank. Run each prompt several times and look at the percentage of runs where you appear.

2. Use a monitoring tool (paid, if you want less manual work)
Several products automate this. Examples mentioned in recent comparisons include OtterlyAI, which starts at $29 a month, Peec AI, Semrush, and Ahrefs Brand Radar. Pricing and coverage change often, so check current details. Most tools report mention rate, position, sentiment, and citations on a recurring basis, which is what you need to spot the first change.

3. Check your analytics
Look for referral traffic from chatgpt.com, perplexity.ai, and similar domains in your analytics tool. This is an indirect signal, and it only shows people who clicked through, but it can catch changes your prompt tests miss.

4. Expect a lag, and know what drives it
Two months is early. AI assistants often rely on what they can find about a product across the web, including reviews, comparison articles, and mentions on sites they trust. A new product may not show up until that coverage exists. Your own website matters less than third-party sources. Several sources in my search also noted that traditional SEO tools do not reveal how often a brand is cited or relied upon in AI-generated answers, so you'll need the prompt testing above.

Practical next step: Run your baseline this week, before you expect any change. Then repeat the same prompts every week or two. Compare the percentage of runs where you appear, not individual answers. That trend will tell you when recommendations start.

To know when AI assistants are recommending your product, you must move beyond traditional search engine tracking and adopt a strategy of Generative Engine Optimization (GEO). Because AI responses are "stochastic" (they change from run to run) and do not provide native analytics dashboards, you need a systematic, repeatable process.

1. How to Track AI Recommendations

Since there is no "rank" to check, you must track patterns rather than isolated answers.

• The Manual "Spot Check" (Free & Immediate):
  • Create a fixed list of 20–40 prompts that your real customers use (e.g., "What are the best [product category] for [use case]?", "Compare [your brand] with [competitor]").
  • Run these queries periodically (e.g., weekly) across the major assistants: ChatGPT, Claude, Perplexity, and Google AI Overviews/Gemini.
  • Document the results in a spreadsheet. Record if you were mentioned, recommended, described accurately, or if a competitor was chosen instead.
• Automated Monitoring Tools:
  • If you want to scale this, several specialized tools now track "AI Visibility" or "AI Share of Voice" by automating these prompt-and-response tests.
  • Look for platforms like Beamtrace, RankZero, Sophyx, or Foglift, which provide dashboards to track your brand’s mention frequency, sentiment, and competitive context over time.
• Analytics & Referral Tracking:
  • Check your web analytics for referral traffic from chatgpt.com, perplexity.ai, or google.com (AI Overviews). While AI traffic is often harder to pinpoint than organic search, some platforms provide referral headers that allow you to isolate these visits.

2. Why AI Might Be Skipping You

If you aren’t appearing, it is rarely due to a "writing trick." AI assistants prioritize trustworthy, structured data. Before you change your marketing copy, audit these technical foundations:

• Check Your "Agent-Readiness": Use tools like Google’s Rich Results Test to see what an AI actually "sees" when it scrapes your site. If your product pages are missing price, stock status, ratings, or clear product names in the structured data, the AI will likely skip you because it cannot verify the information.
• Remove Technical Blocks: Ensure your robots.txt file is not blocking crawlers like GPTBot, ClaudeBot, or Google-Extended.
• Corroboration is Key: AI models prioritize information confirmed by multiple third-party sources. Getting your product mentioned on reputable industry blogs, review sites, and directories helps the AI "build confidence" in your brand.

3. The "Visibility" Framework

When you monitor your results, classify them into one of these stages to know how to improve:

• Invisible: The AI doesn't know you exist. Fix: Improve structured data and site indexing.
• Cited: The AI links to your site but doesn't recommend you. Fix: Strengthen your brand’s presence on trusted third-party sites.
• Recognized: It knows you when asked directly but ignores you in discovery. Fix: Improve your product's category associations and descriptions.
• Recommended: You are actively suggested. Goal: Maintain by ensuring data consistency and positive customer feedback.

Pro Tip: Don't panic if you see a competitor in one result and you in another. Because AI answers are probabilistic, trends over weeks are the only data that matters.

You won’t get a single reliable notification that an assistant has “started recommending” your product. Treat it as something to measure over time across the assistants your customers use.

1. Make a small prompt list. Write 10–30 questions buyers might ask, such as “What’s a good [product type] for [use case]?” or “Alternatives to [competitor]?” Include your product name in only some prompts.
2. Check the same prompts regularly. Test them in ChatGPT, Gemini, Perplexity, and other relevant assistants, using the same country, language, and account conditions where possible. Record whether your product is named, described positively, linked or cited, and which competitors appear. A mention in one answer is a clue—not proof of broad visibility.
3. Track the trend. Calculate the share of prompts where your product is recommended, and compare it month to month. Also save the exact answers and cited sources so you can see what changed. AI visibility tools can automate this across prompts and platforms, but compare their coverage and methodology before choosing one. ⁶⁸
4. Watch for traffic and demand. Check analytics for visits referred from AI services and for later increases in branded searches or product-page visits. These are useful signals, but they won’t capture every recommendation: someone may see your product in an answer and never click. Google Search Console includes AI-feature activity in its search reporting, though those metrics don’t show what every chatbot says about your brand. ²⁵
5. Improve the evidence assistants can find. Keep your product pages clear and specific, and build a consistent footprint in trustworthy reviews, comparisons, directories, and other relevant sources. When you see an assistant citing a page, note whether it’s yours or someone else’s.

For a new product, start with a simple spreadsheet and a weekly or biweekly check. You’ll learn more from a consistent set of realistic prompts over several weeks than from occasional spot checks.