How to get a Base44 app cited in ChatGPT, Claude and Perplexity answers

AI assistants cite pages they can read, and on Base44 most of the important AI crawlers can read your pages once Base44 has rendered them. GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, PerplexityBot and Perplexity-User all got Base44’s pre-rendered page in our tests. The gaps are pages Base44 hasn’t rendered yet, which get a generated summary, and a set of crawlers Base44 doesn’t render for at all. After that, it comes down to the same things as search: pages that rank, facts stated plainly, and a clear brand.

This builds on the Base44 SEO guide. Here’s what’s specific to AI answers.

How AI assistants find your pages#

There are two routes, and they need different things.

1. Training. Crawlers such as GPTBot, ClaudeBot and CCBot collect pages that may end up in a model’s training data. That’s slow, and you can’t steer it much.
2. Live retrieval. When someone asks a question, assistants search an index and fetch pages to answer it. OAI-SearchBot and ChatGPT-User (OpenAI), Claude-User (Anthropic) and Perplexity-User (Perplexity) are the fetchers behind this. It’s where citations with links come from.

Base44’s SEO docs make a point worth repeating: “Optimizing your app for search engines also improves its visibility to AI assistants that rely on those search indexes.” If a page doesn’t rank anywhere, live retrieval is unlikely to find it.

Which AI crawlers get the rendered page#

Base44 decides by user agent. In our test of 47 Base44 sites, we requested pages that already had a render with each crawler’s user agent.

“Summary only” means the crawler got Base44’s hidden, auto-generated block (seo-snapshot): a heading made from the page and app name, a stock description of your app, and a list of links. It tells a model that your app exists and roughly what it does. It doesn’t contain your page’s copy, prices, features or answers.

Two of those rows deserve a second look.

• CCBot builds Common Crawl, a public web archive that has been widely used in training datasets. On Base44, it collects summaries.
• Bravebot builds Brave Search’s independent index, which other products license for web results. Brave doesn’t use Google or Bing, so it has to crawl you itself, and on Base44 it gets the summary.

llms.txt: switch it on, expect little#

Base44 can generate an llms.txt file, “a lightweight file that gives AI crawlers like GPTBot, ChatGPT-User, ClaudeBot, and PerplexityBot a clear, structured summary of what your app does.” It’s off by default. Switch it on from the SEO & GEO page’s Setup checklist.

It’s free and harmless, so turn it on. Keep expectations low: there’s little public evidence that the major assistants read llms.txt when answering questions. Pages that crawlers can read in full matter far more.

curl -s https://yourdomain.com/llms.txt | head -20

What makes a page citable#

Assistants cite pages that answer the question directly and state facts they can repeat. Base44’s own guidance lines up with that, and it’s worth following:

• Answer in the first two sentences. A page that opens with a tagline gives an assistant nothing to quote.
• State specific, checkable facts. Base44’s docs suggest “at least 2 specific numbers or statistics” in your app description. Prices, sizes, timeframes, locations and counts all get quoted.
• Use your customers’ words. Headings phrased as the questions people actually ask.
• Mark up FAQs. A plain FAQ section, wrapped in FAQPage structured data. The structured data guide covers it.
• Make the business real. Business name, address, team, policies. Assistants recommending a business want to know it exists.
• Keep your name, favicon and description consistent. Base44’s docs note that in AI-generated lists, “the site name and favicon are often the only brand assets someone sees.”

Tracking whether it’s working#

On the Builder plan and above, Base44’s SEO & GEO page shows AI Visibility metrics over seven days: AI crawls, reads in AI answers, visits from AI, and your most-read page. It counts fetches. It can’t show you what each fetch received.

For a direct check, ask the assistants yourself. Put the questions your customers ask into ChatGPT, Claude and Perplexity every few weeks and note whether your app is mentioned and linked. Results vary by phrasing and by session, so track a fixed set of questions over time. If you want this done on a schedule across several assistants, tools like Encited track brand mentions in AI answers and show which crawler fetched which page.

Should you block AI crawlers?#

Base44’s robots.txt allows AI crawlers by default, and for most apps that’s the right call. You can’t appear in an assistant’s answers while blocking the crawler that fetches for it.

If your content is the product, such as paid research or a dataset, blocking training crawlers like GPTBot and CCBot while allowing live fetchers is a reasonable middle ground. Base44 doesn’t document a way to edit its generated robots.txt, so check with support before planning around it.

What to do next#

1. Turn on llms.txt.
2. Rewrite the first two sentences of your key pages so they answer a question.
3. Add FAQPage markup where you have FAQs.
4. Check with GPTBot’s user agent that your key pages return the rendered page:

UA='Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; GPTBot/1.2; +https://openai.com/gptbot'
curl -s -A "$UA" https://yourdomain.com/your-page | grep -c 'id="seo-snapshot"'
# 0 = rendered page, 1 = generated summary