A 6-Step Checklist for Product Discovery in the AI Era

When your friends have all signed up, the next wave of users needs to find you on their own—and that mostly happens on Google and in AI assistants. The Serpair team has put together the "discoverability" technical checklist they most wanted after launch, ordered by execution sequence. The core idea comes from YC's Ankit Gupta: finding your first users is more of a search problem than a persuasion problem.

1. First, make sure your own name can be searched

If someone hears your product name and searches for it, you should rank first. Many new products fail at this, especially when the name is made up of common words or is already taken by another company.

How to check: search your product name in a private window, then search "name + category"; verify your site in Google Search Console and use URL Inspection to confirm your homepage is indexed; search your name in ChatGPT, Perplexity, and Gemini to see what they think you are.

How to fix: put both the name and the category into your homepage title and first H1—"Acme: Invoicing for Freelancers" is far better than "Acme" alone; use exactly the same name across your site, social profiles, launch platforms, and GitHub; use WebSite structured data on your homepage to tell Google your preferred site name, and optionally add alternateName as an alternative. Google's official documentation explicitly states that this markup is "the most important" if you want to declare a preference.

2. Add Organization structured data, but don't expect too much

Organization markup helps Google "disambiguate" your organization in search results, which matters especially when something else shares your name. The sameAs property links your official profiles together. Every sameAs URL must be manually verified—one misspelled handle will point Google at someone else.

A word of caution here: structured data is not a shortcut to earning AI citations. Ahrefs tracked 1,885 pages with JSON-LD added and found no significant increase in their citations on Google AI Mode or ChatGPT. However, those pages were already heavily cited to begin with, so the conclusion is of limited relevance for brand-new sites. Google also states that its AI features don't require "special schema.org structured data," and that structured data should match the visible text. So: add schema to help Google figure out who you are, but the facts that really matter need to be written into the body text users can actually see.

A brief introduction to structured data (Schema.org) and JSON-LD

Structured data is a mechanism for tagging web content using a shared vocabulary (schema.org, co-founded by Google, Microsoft, Yahoo, and Yandex) so that search engines don't have to guess a page's meaning from natural language alone. Common types include Organization, WebSite, Product, FAQPage, and so on. JSON-LD is the serialization format Google recommends, typically embedded in <head> or <body> via <script type="application/ld+json">, kept separate from visible text and easy to maintain.

sameAs is a key property under Organization, used to declare the organization's official profiles on other platforms (such as LinkedIn, X, GitHub), helping Google merge these scattered identities into a single entity. alternateName is used to declare aliases or abbreviations for the entity. It's worth emphasizing: the facts described by structured data must be consistent with the page's visible content—otherwise it may be considered misleading and won't bring ranking or citation benefits.

3. Let AI crawlers in

AI assistants can only cite pages their crawlers can reach. Each has its own documentation:

• OAI-SearchBot powers ChatGPT search; OpenAI says sites that opt out "will not appear in ChatGPT search answers";
• GPTBot is OpenAI's training crawler, a separate setting from search, so you can block training while keeping search;
• PerplexityBot is responsible for surfacing and linking sites in Perplexity results;
• Googlebot covers Google Search, including AI Overviews and AI Mode; Google says robots.txt rules targeting Googlebot are the control switch for these features.

A minimal robots.txt that welcomes search crawlers should explicitly allow the user agents above and declare a sitemap. But robots.txt isn't the only gate: your CDN or firewall may also block them. Google lists CDN and hosting settings as items to check, and Perplexity's documentation includes rules for allowing its crawler through Cloudflare and AWS WAF.

You can run a smoke test from the terminal, requesting your homepage with the user agent from OpenAI's documentation:

curl -s -o /dev/null -w "%{http_code}\n" \
  -A "Mozilla/5.0 (compatible; OAI-SearchBot/1.4; +https://openai.com/searchbot)" \
  https://acme.com/

A 403 or a challenge page means something is blocking the bot. Some firewalls also verify real IP ranges (both OpenAI and Perplexity publish theirs), so a 200 is only a good sign, not proof.

robots.txt and user-agent matching

robots.txt is a plain text file placed at your site's root that tells crawlers which paths can be fetched through User-agent and Disallow/Allow directives. Before requesting each URL, a crawler reads this file and matches the rule group corresponding to its user-agent string; crawlers that match no group usually follow the general rules under User-agent: *.

Therefore, the correct way to "allow a given AI crawler" is to create a separate group for its user agent with Allow: /, rather than relying only on the general rules—many sites write Disallow: / in the general group, which will also block crawlers that aren't explicitly listed. Note that robots.txt only constrains well-behaved crawlers and does not constitute access control; actual blocking often comes from CDN or WAF bot management rules, which is why a curl smoke test is needed later.

4. Write a page that actually answers a question

Being crawlable doesn't mean being useful. AI search tools often issue multiple related searches for a single question—Google calls this "query fan-out"—and then pull snippets from pages that clearly answer a specific sub-question.

What to do: pin down the specific problem you solve using users' own words (look on Reddit, Hacker News, and in support emails for how they describe it); write a page that gives the answer in the first two or three sentences before expanding into details; keep important content in HTML text, not only in images, videos, or client-side components; add a short FAQ collecting the questions people actually follow up with.

What query fan-out means

Query fan-out refers to how AI search systems, when answering a user's question, don't issue just one retrieval—they break the original question into multiple related sub-queries issued in parallel, then synthesize snippets from various sources to generate an answer. For example, if a user asks "what invoicing tool should a small team use," the system might simultaneously search "invoicing tool recommendations for small teams," "invoicing tool pricing comparison," "invoicing software for freelancers," and so on.

This means your page doesn't need to cover the entire broad topic; it should precisely hit one specific sub-question and give an answer sentence that can be excerpted right at the start. Clear page structure, explicit subheadings, and answers up front make it more likely that one of the fan-out retrievals will select and cite you.

5. Show up in the sources behind the answers

AI answers favor sources the web already trusts. Pew's research found that Wikipedia, YouTube, and Reddit are the three most-cited sources in Google AI summaries. You can't buy your way in, but you can appear honestly: list your product in relevant directories and launch sites with a consistent name and one-line description; answer real questions in the communities where users ask, mentioning your product only when it's genuinely helpful; write comparison or "how-to" articles on platforms and link back to your site.

6. Measure, then repeat

In GA4 or your analytics tool, look at sources like chatgpt.com, perplexity.ai, and gemini.google.com. Some AI visits have no referrer and land in direct traffic, so this number is only a lower bound. In Search Console, watch impressions for your brand terms and question phrases. Re-run the same prompts every few weeks—SparkToro found there's less than a one percent chance that an AI tool gives the same brand list, so track how often you appear across multiple runs rather than your rank in a single answer.

Serpair itself offers this kind of tracking: monitoring Google rankings and whether ChatGPT, Perplexity, and Gemini mention you, then turning that into a short fix list.

Your friends helped you launch; this checklist helps the next wave of people find you.

One-line Summary

Finding your first users for a new product is essentially a search problem—follow this six-step checklist to get found, crawled, and cited.

Paragraph Summary

Drawing on YC's Ankit Gupta's view that finding your first users is more of a search problem than a persuasion problem, the article presents a discoverability checklist ordered by execution sequence: first ensure your product name can be searched and ranks first, using WebSite structured data to declare your site name; add Organization structured data to help Google disambiguate, while noting that structured data is not a shortcut to AI citations; allow AI crawlers like OAI-SearchBot and PerplexityBot at both the robots.txt and CDN/firewall layers, with a curl smoke test; given the query fan-out nature of AI search, write pages that answer up front; appear in trusted sources via directories, communities, and comparison articles; and finally measure with analytics tools and Search Console, re-running prompts regularly to track how often you appear.