How to get your startup mentioned by ChatGPT and Perplexity

GEO starts with a list, not with content. Write down the questions a buyer types into ChatGPT or Perplexity before they would ever find you. Most fall into five shapes:

1. "Best X for Y": "best scheduling app for small restaurants"
2. "X vs Y": "7shifts vs Homebase"
3. "X alternatives": "cheaper alternatives to 7shifts"
4. "How do I…": "how do I make a staff schedule for a diner"
5. "Is X worth it": "is 7shifts worth it for a 10-person team"

The examples above are for Shiftly, the fictional restaurant scheduling app we use across this site. Do the same for your business. Aim for 20 questions, with at least five comparisons that name your real competitors.

Where to find them:

• Sales calls and support emails. The words customers use when they describe the problem.
• Search Console. Queries you already get impressions for, especially long ones.
• Reddit and community threads. Search your category plus "recommend" or "alternative".
• The assistants themselves. Ask ChatGPT "what would someone ask before buying a [category]?" and keep the realistic ones.

Keep the list in a spreadsheet. It becomes your measurement set and your content plan.

Before you write anything, ask every question in ChatGPT, Perplexity and Gemini and write down what comes back. For each answer, record:

Two things to know:

• Answers vary. The same question can return a different list an hour later. One check means little. A weekly check over a month means a lot.
• Sources are a map. When Perplexity cites a G2 page, a Reddit thread and a competitor's comparison article for a question, those are the places you need to appear.

Turn the results into one number: the share of answers that name you. That's your share of voice in AI answers. If you're at 0%, you're in good company. Most new products are.

Assistants answer questions by pulling from pages that already answer them clearly. So write those pages, one per question cluster.

Answer in the first two sentences. If the question is "best scheduling app for diners", the page should open with the answer and the reason, not with a story. For example:

The best scheduling apps for small diners are Shiftly for single-location teams under 20 that want shift swaps by text, and Rotaly for multi-location groups that need payroll built in. Both have free trials; Shiftly is cheaper for one location.

(Both products in that example are made up.) That paragraph can be lifted straight into an answer. A page that opens with "In today's fast-paced restaurant industry…" can't.

The pages that work best:

• "X vs Y" pages. One per real competitor. A table, a verdict, and "choose them if / choose us if". Be fair. Name their real strengths. Assistants and readers both trust pages that admit trade-offs.
• "X alternatives" pages. You first, plus two to four honest other options with who each is for.
• "X for Y" pages. One per industry or use case you actually serve, with specifics for that industry.
• How-to guides that answer the job, not just the product.

Make every page specific. Numbers, prices with a "checked on" date, named integrations, real limits. Vague pages give the model nothing to quote.

Date your facts. "Pro is $69 a month (checked November 2026)" is more useful than "affordable pricing", and it shows the page is maintained.

Your own site is one source. Assistants also read what other sites say about you, and many weigh those more. Use the cited-pages column from your measurement sheet to decide where to show up.

Places that usually matter:

• Launch directories. Product Hunt, BetaList, There's An AI For That, SaaSHub, AlternativeTo and dozens more. Each listing is a page that says what you do. See launch on 50 directories in a week.
• Review sites. G2, Capterra and GetApp pages often appear as sources for "best X" questions. A handful of real reviews beats an empty profile.
• Community threads. Reddit, Hacker News and Indie Hackers threads where people ask for recommendations. Answer helpfully, say you're the maker, and post from your own account. Don't automate this; it gets accounts banned.
• Other people's comparison articles. If a blog ranks for "best scheduling apps", ask the author to look at you. Offer a trial and facts, not a payment.
• Podcasts, newsletters and guest posts. Show notes and articles are more pages that name you in context.

The goal isn't volume. It's being named, correctly, on the pages that are already sources for your questions.

Assistants and the crawlers behind them read HTML. Make the important facts easy to extract.

• Schema markup. Add JSON-LD: Organization on the home page, SoftwareApplication with an Offer (price and currency) on pricing and product pages, FAQPage where you have an FAQ, BreadcrumbList on inner pages, and Article with a modified date on guides.
• llms.txt. A short Markdown summary of your site at /llms.txt: what you are, the key facts, and links to your main pages with one line each. See what llms.txt is. Ours is at /llms.txt.
• A clean sitemap that lists every page you want found, with real modified dates.
• robots.txt. Check you aren't blocking the crawlers you want to be read by. ChatGPT search uses OAI-SearchBot and Perplexity uses PerplexityBot. Decide deliberately rather than by accident.
• Fast, plain HTML. If the main text only appears after heavy JavaScript runs, some crawlers will see an empty page.

None of this gets you named on its own. It removes reasons not to be.

If your site says $99 a month, your Product Hunt page says $79 and an old review says "free", an assistant has to pick one or hedge. Often it hedges, or names a competitor whose facts agree.

Write one fact sheet and use it everywhere:

• one-line description (under 20 words),
• who it's for,
• price and plan names,
• main integrations or channels,
• founding year and founder name, if you share them.

Then update every listing you control: your site, directories, review profiles, LinkedIn, X, Crunchbase. When the price changes, change it everywhere in the same week.

Here's a realistic first month for a small team. It assumes a few hours a week, or an AI CMO doing most of it.

Week 1: measure and fix the basics

• Write the 20 questions and run the baseline in three assistants.
• Write the fact sheet. Fix your homepage one-liner and pricing page to match it.
• Add Organization, SoftwareApplication and FAQPage schema. Publish llms.txt. Check robots.txt.

Week 2: the comparison pages

• Publish "vs" pages for your top three competitors, fair and specific.
• Publish one "alternatives" page for your biggest competitor.

Week 3: mentions

• Submit to the first 20 launch directories with a specific listing for each.
• Claim your G2 or Capterra profile and ask five happy customers for a review.
• Answer two real community threads, as yourself.

Week 4: industry pages and re-measure

• Publish "for [industry]" pages for your two best-fit customer types.
• Re-run the 20 questions. Compare with week 1. Note which pages got cited.

After that, keep the weekly measurement going and add pages where competitors dominate.

GEO is part of Cosmo's SEO + GEO module, in the one $99 plan. He builds the question list from your site and competitors, runs the weekly check across ChatGPT, Perplexity and Gemini, and writes the comparison, alternatives and industry articles. He publishes them to WordPress, Webflow, Shopify, Ghost, Wix, GitHub or Framer after your 👍, writes your llms.txt, and submits you to directories.

The morning brief tells you when something moves: "Perplexity now names you in 3 of 10 'best scheduling app for diners' answers. Last month it was 0." Nothing goes out without your yes. More on what an AI CMO does.