The Base44 SEO & GEO score: what it checks, what it fixes, what it misses

Base44’s SEO & GEO scan checks whether your app is set up the way search engines and AI assistants expect: titles, descriptions, structured data, crawl files and content signals. It’s a useful checklist, and Fix with AI clears most of it quickly. What it can’t tell you is whether a crawler actually received your page on its last visit, and on Base44 that’s the question that matters most.

Here’s what the scan covers, what to do with each finding, and the checks to run yourself. It sits inside the wider Base44 SEO guide.

Where to find it and what it shows#

Base44 launched the SEO & GEO tab on 26 April 2026, under Marketing in your app dashboard. Per Base44’s documentation, a scan gives you:

• A score as a percentage, with a label such as “Moderate” or “Good.”
• Urgent issues, “directly hurting your visibility and should be fixed first.”
• Recommended improvements, “that can boost your score but are not blocking.”
• Done, the checks you already pass.

The same page holds the Meta tags tab for per-page titles and descriptions, the Setup checklist (including the sitemap and llms.txt toggles) and Advanced Settings (per-page indexing and the injection toggles).

Publish first, then scan#

The docs are clear that an unpublished app can’t score well: “Search engines and AI tools can only find a published app, so until you publish, your score cannot rise above a certain point.” Publish, connect your custom domain, then scan. Base44 calls the custom domain “the single biggest improvement” you can make.

What Fix with AI does#

Fix with AI sends every AI-fixable item to the chat composer so the AI applies them in one pass. It uses message credits like any other chat request. Some fixes apply directly without the AI, including auto-generated breadcrumbs, auto-generated brand info and improved descriptions.

Things the scan can set up for you:

• Titles and descriptions per page.
• WebSite, Organization and FAQPage structured data, which Base44 then injects. BreadcrumbList is generated for every route regardless.
• llms.txt, once you switch it on.

Things it asks you for: original messaging, up-to-date facts, and anything that needs knowledge of your business.

What the score doesn’t measure#

A high score says your configuration is right. Three things it can’t see:

1. What crawlers received. Base44 serves recognized crawlers a cached, pre-rendered copy of each page. When a page has no cached render, crawlers get a hidden, auto-generated summary block instead of your page. In our test of 47 live Base44 sites, a first Googlebot request got the real page on only 34% of inner pages. The scan doesn’t report this. Test it yourself:

UA='Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)'
curl -s -A "$UA" https://yourdomain.com/your-page | grep -c 'id="seo-snapshot"'
# 0 = rendered page, 1 = generated summary

2. Record pages. The scan works per page file. A template page that shows a thousand products counts as one page with one title, and none of the products are in your sitemap.

3. Whether the content is worth ranking. Structure, tags and schema make a page eligible. What earns the ranking is still the content and the links pointing at it.

AI Visibility metrics#

On the Builder plan and above, Base44 adds AI Visibility metrics over a rolling seven days:

• AI crawled your app, counting fetches by AI bots.
• Your app in AI answers, when an assistant reads your app for a specific query.
• Visits from AI, referral clicks from assistants.
• Most-read page.

Base44 notes that “assistants never share what the person actually typed”, so you’ll see that a read happened without the question behind it. The metrics count fetches. They don’t tell you whether a given fetch got the rendered page or the summary. For that you need per-request logs from a proxy in front of your domain, or the curl test above.

A sensible order of work#

1. Publish on your custom domain and run the scan.
2. Fix everything under Urgent.
3. Run Fix with AI for the rest, then rewrite titles and descriptions on your top pages by hand.
4. Turn on llms.txt.
5. In Advanced Settings, turn off indexing for login, password-reset and admin pages. The scan won’t necessarily flag them, and in our test they were 21% of sitemap URLs.
6. Run the curl test on your key pages after each publish.

What to do next#

Treat a score of 90% or more as done for configuration and move on to content, record pages and the meta tags the scan can’t write for you. If you care about AI answers, the AI search guide picks up from here.