We checked 313 new SaaS websites. Three in four have the basics

On this page

1. What did we check?
2. Do new SaaS sites have the technical basics?
3. How many new SaaS sites show crawlers an empty page?
4. Does having an llms.txt file help?
5. Is FAQ markup still worth adding?
6. What about software markup and ratings?
7. Do new SaaS sites block AI crawlers?
8. Where are the gaps?
9. What should you do with this?
10. Method and limits
11. Questions
12. Sources

Three in four newly launched SaaS websites already have the technical basics in place. We checked 313 sites listed on eight launch directories in 2026. 233 of them (74.4%) had readable text, a meta description, a canonical tag, a sitemap, a robots.txt file, and structured data. Roughly one in eleven showed crawlers an empty page.

What did we check?

We loaded each site the way a crawler that does not run JavaScript would. That means one request for the home page, then robots.txt, the sitemap, and llms.txt. No browser, no scripts, no logins.

For each site we recorded what was in the HTML the server sent back. That covered how many words a reader could see, whether there was a title, a meta description, and a canonical tag, and what structured data was present.

We also noted which crawlers robots.txt named or blocked, how many pages the sitemap listed, and what kind of pages they were.

The sample was 320 products taken from eight launch directories, 40 from each. 313 could be analysed, and every figure marked as ours in this article comes from that check⁴. Nine in ten were listed in September 2026.

The full method is at the end of this article, along with its limits.

Do new SaaS sites have the technical basics?

Most do. We picked six basics that decide whether a page can be found and read at all, and 233 of 313 sites had every one.

Every one of the 313 had a title tag. 29 sites (9.3%) were missing three or more of the six.

These new sites are ahead of the web at large. The HTTP Archive's Web Almanac, which crawls millions of sites, found canonical tags on 68% of desktop pages, meta descriptions on 67.7%, and JSON-LD on 43% of home pages in its 2025 edition¹⁰. The two sets of numbers were gathered in different ways, a year apart, so treat the comparison as a rough guide.

Our reading is that the tools founders now build with produce these files by default. We did not test that, so take it as an opinion.

The numbers do show that a sitemap and a tidy set of meta tags no longer set a new product apart. Most of the products launching beside yours have them too.

How many new SaaS sites show crawlers an empty page?

29 of 313 new SaaS home pages, or 9.3%, showed crawlers an empty page. On those sites the home page had fewer than 50 readable words in the HTML the server sent. On 26 of them the count was zero.

An empty page in this sense still looks fine in a browser. The server sends a near-blank file and a script, and the script builds the page on the visitor's device. A crawler that does not run the script gets the blank file.

Most of these sites had done the rest of the work. 27 of the 29 had a meta description. 19 had a sitemap. 13 had structured data.

The text of the page was the missing part.

How the site was built made a large difference.

One in three of the Vite single page apps sent an empty page. Vite itself is a build tool and can produce pages that work without JavaScript. The sites in this group were set up to render in the browser only.

This matters because AI crawlers don't run JavaScript, as far as the published evidence goes. Vercel and MERJ studied crawler traffic in late 2024 and reported that "none of the major AI crawlers currently render JavaScript"¹.

Anthropic's developer documentation says its web fetch tool "does not support websites dynamically rendered with JavaScript"³. Google does render JavaScript, and its own documentation still recommends server side rendering or pre-rendering because "not all bots can run JavaScript"².

We cover how to test your own site, and the fixes, in Can Google and AI assistants read a site built with AI tools?.

Does having an llms.txt file help?

The evidence says no, and yet more than half of these sites have one. 167 of 313 (53.4%) had an llms.txt file at the root of the site. The file is a proposal, first published in September 2024, for a plain text guide to a site written for AI tools¹⁴.

That is far above the wider web. The Web Almanac found a valid llms.txt on 2.13% of desktop sites in 2025¹⁰. Ahrefs found the file on 28% of the 137,210 domains it analysed in 2026, a sample that leans towards marketing-aware sites⁷.

Ahrefs also looked at whether anything reads these files. 97% of them received no requests at all in May 2026⁷. Google's guide to generative AI features says llms.txt "will neither harm nor help your site's visibility or rankings in Google Search"⁸.

The file costs little and does no harm. It is also the most adopted item in our check that has no measured benefit.

Is FAQ markup still worth adding?

For Google's search results, no. 143 of 313 home pages (45.7%) carried FAQPage structured data. Google's documentation changelog records that "the FAQ rich result feature is no longer shown in Google Search results", with the feature ending on 7 May 2026⁹.

Nearly half of the sites launching this month are carrying markup for a feature that was switched off four months earlier. Google has not said the markup does harm.

A questions section that visitors can read is still useful on the page. The markup around it no longer earns anything in Google's results.

101 sites (32.3%) had both an llms.txt file and FAQPage markup.

What about software markup and ratings?

173 of 313 home pages (55.3%) carried SoftwareApplication or WebApplication structured data. Google supports this type for rich results, with conditions. To be eligible a page must include the app name, an offer price, and either an aggregate rating or a review¹¹.

19 of those 173 sites (11.0%) included an aggregate rating. 18 of the 19 stated a rating of 4.5 or higher, and 7 stated a count of 100 ratings or more.

We did not check whether those ratings exist, so we draw no conclusion about any site.

A product in its launch week rarely has the reviews that this markup is built to describe. Markup should only describe reviews that exist.

Do new SaaS sites block AI crawlers?

Hardly any. 8 of 313 sites (2.6%) blocked at least one AI crawler from the whole site in robots.txt. GPTBot was blocked by 4, ClaudeBot by 2, and PerplexityBot by none.

94 sites (30.0%) named at least one AI crawler in robots.txt, and in nearly every case the rule was there to allow it. GPTBot and PerplexityBot were each named by 85 sites, and ClaudeBot by 81.

For comparison, the Web Almanac found GPTBot named in the robots.txt of 4.5% of desktop sites in 2025¹⁰. Among the top 10,000 domains, Cloudflare found that about 14% of those with a robots.txt had rules aimed at AI bots, and GPTBot was the one most often disallowed¹².

The biggest sites are the ones most often shutting AI crawlers out, whilst the new SaaS products in our check are letting them in.

Where are the gaps?

The gaps are in the content. The technical side is mostly in place, and the pages that do the selling often are not.

The last row is the one to look at. 48 sites had done everything on the technical list and had no blog and no comparison page.

This is where the evidence on AI visibility points. In an Ahrefs study of 75,000 brands, mentions of a brand across the web correlated with AI visibility far more strongly than backlinks did⁵.

When Ahrefs tested structured data directly, by tracking 1,885 pages that added it, "adding schema produced no major uplift in citations on any platform"⁶. Google's own guidance for AI search asks for unique content and pages it can access⁸.

Buyers are moving in the same direction. In a G2 survey of 1,076 B2B software buyers in March 2026, 51% said they start software research with an AI chatbot more often than with Google, up from 29% a year earlier¹³. G2 is a review site with an interest in that finding, so read it with that in mind.

We look at what the studies say in How AI assistants choose which products to recommend, and at the pages most new sites are missing in The comparison and alternatives pages every SaaS needs.

What should you do with this?

Start by finding out which group you're in.

1. Open your home page, view the page source, and search it for a sentence you can see on the page. If the sentence is missing from the source, crawlers that don't run JavaScript can't read it.
2. Check the six basics in the first table. Each one takes minutes to confirm.
3. If all six are present, stop working on the technical side. Put the time into pages that answer what buyers ask.
4. Leave llms.txt and FAQ markup alone. Keep them if you have them, and don't spend time adding them.

A week by week version of this is in What to fix on your SaaS website in the first week after launch.

Method and limits

Sample. 320 products from eight launch directories: BetaList, Launching Next, Uneed, Microlaunch, TinyLaunch, Fazier, DevHunt, and StartupBase. 40 from each, chosen in random order with a fixed seed. Products were excluded if they were mobile only, agencies, directories, courses, physical goods, or had no website of their own.

Listing dates ran from 24 February to 28 September 2026. Of the 309 analysed sites with a listing date, 288 were listed in September 2026.

Check. Run on 28 September 2026. Of the 320 sites, 6 could not be reached and 1 redirected to a different domain, leaving 313. Each site was requested with a desktop browser user agent that also identified our checker. JavaScript was not run.

Definitions.

• Readable words: words in the HTML after removing scripts, styles, and markup.
• Empty page: fewer than 50 readable words on the home page.
• Comparison or alternatives page: a URL in the sitemap containing "vs", "compare", "comparison", or "alternative".
• Blog: a link on the home page to a blog, articles, or resources section, or matching pages in the sitemap.
• Build marker: a pattern in the HTML associated with a framework or tool.

Limits.

• Launch directories are self-submitted. The sample leans towards solo founders and AI tools, and it is not a sample of all new SaaS products.
• A listing date is the date a product appeared in a directory. We did not check how old each product is.
• We checked the home page only. A site can have a full home page and empty inner pages, or the reverse.
• We recorded presence only. A meta description counted if it existed, and we made no judgement of its quality.
• The 50 word threshold is our own choice. Counting sites under 200 words gives 40 of 313 (12.8%).
• URL patterns miss pages with unusual addresses and can catch pages that are not comparisons.
• Sites can respond differently to different user agents. An AI crawler may be served something other than what we saw.
• No site is named. Results are published as totals.

The totals are available as a CSV file, and as a table on the research page.

Questions

How many new SaaS websites were checked?

320 were sampled and 313 were analysed. The check ran on 28 September 2026 across products listed on eight launch directories.

What share of new SaaS websites can't be read without JavaScript?

9.3%. 29 of 313 home pages had fewer than 50 readable words in the HTML, and 26 of those had none.

Which way of building a site had the most empty pages?

Sites with the markers of a Vite single page app. 18 of 54 (33.3%) had fewer than 50 readable words, against 1 of 134 (0.7%) for Next.js.

How many new SaaS websites have an llms.txt file?

53.4%, which is 167 of 313. Ahrefs found that 97% of llms.txt files received no requests in May 2026, and Google says the file does not affect visibility in Google Search.

Do new SaaS websites block AI crawlers?

Very few. 2.6% blocked at least one AI crawler from the whole site. 30.0% named at least one AI crawler in robots.txt, almost always to allow it.

What are new SaaS websites most often missing?

Content. 40.6% showed no blog or articles section, 46.6% had no comparison or alternatives page, and 36.1% had no pricing page we could find.

Sources

1. Vercel and MERJ, "The rise of the AI crawler", 17 December 2024. https://vercel.com/blog/the-rise-of-the-ai-crawler
2. Google Search Central, "Understand the JavaScript SEO basics", updated 4 March 2026. https://developers.google.com/search/docs/crawling-indexing/javascript/javascript-seo-basics
3. Anthropic, Claude developer documentation, "Web fetch tool". https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-fetch-tool
4. SaaS GTM, site check of 313 newly launched SaaS websites, 28 September 2026. Own data.
5. Ahrefs, "Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied)", 12 December 2025. https://ahrefs.com/blog/ai-brand-visibility-correlations/
6. Ahrefs, "We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved.", 11 May 2026. https://ahrefs.com/blog/schema-ai-citations/
7. Ahrefs, "We Analyzed 137K Sites: 97% of llms.txt Files Never Get Read", 15 June 2026. https://ahrefs.com/blog/llmstxt-study/
8. Google Search Central, "Optimizing your website for generative AI features on Google Search", updated 10 July 2026. https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
9. Google Search Central, "Latest documentation updates", entry on removing the FAQ rich result, updated 24 September 2026. https://developers.google.com/search/updates
10. HTTP Archive, Web Almanac 2025, chapter 5, "SEO", published 15 January 2026. https://almanac.httparchive.org/en/2025/seo
11. Google Search Central, "Software app (SoftwareApplication) structured data", updated 8 September 2026. https://developers.google.com/search/docs/appearance/structured-data/software-app
12. Cloudflare, "From Googlebot to GPTBot: who's crawling your site in 2025", 1 July 2025. https://blog.cloudflare.com/from-googlebot-to-gptbot-whos-crawling-your-site-in-2025/
13. G2, "New G2 Research: Half of B2B Software Buyers Now Start Their Research With AI Chatbots", 15 April 2026. https://www.prnewswire.com/news-releases/new-g2-research-half-of-b2b-software-buyers-now-start-their-research-with-ai-chatbots-302742807.html
14. llmstxt.org, "The /llms.txt file", first published 3 September 2024. https://llmstxt.org/