B2B SaaS SEO is search optimization for a software company’s own site, aimed at the business buyers who compare tools before they book a demo. The nine pages ranking on Google for “b2b saas seo” in September 2026 included eight written by SEO agencies, three of them service pages. The AI answer on the same search told software companies four specific things to do, and none of the ten pages we read had checked whether the software companies people have heard of do them. So on 28 September 2026 we checked 69 of them. This page covers what that advice says, where the blogs actually live, whether the pages arrive without JavaScript, the pages built for buyers comparing tools, the schema, and what the sites tell AI crawlers and agents.
What the advice on this search says
When we captured the search in September, the AI panel above the results was an AI Mode reply, in the slot where an AI Overview usually sits. It told a SaaS company four things, each credited to a source:
- Host the blog “on a clean subfolder … rather than a separate subdomain to preserve domain authority”.
- “Maintain fast site speeds, proper schema markup, and clean JavaScript rendering so both traditional crawlers and AI bots can index your pages.”
- Target “competitor comparisons” and “best X software” lists.
- “Optimize content with precise citations and structured data” for ChatGPT and Perplexity.
The first source the answer cited was a press release from an SEO agency, distributed through GlobeNewswire and republished on Yahoo Finance. Three of its six sources did not appear in the organic results below it.
All four pieces of advice can be checked on a live site, so we checked them.
What we measured, and how
The sample is the 71 software and SaaS companies from our study of AI crawler rules: recognisable, mostly large companies such as Stripe, Notion, HubSpot, Slack and Vercel. On 28 September 2026, 69 of their homepages could be read. One did not answer, and one served our crawler a page asking it to update its browser, so both are left out.
Each homepage went through the same checks as our free audit tool. Then we found the blog, from the homepage’s own links or a request for /blog; read the pricing page; read the sitemap for pages comparing the product with other tools; and asked each homepage and pricing page for Markdown, the way an AI agent can. We read 68 sitemaps and 67 pricing pages.
Everything was read from the delivered HTML, the way AI retrieval crawlers read a page. The requests went out from India, so three sites served their Indian editions. The pattern search for comparison pages was checked by hand: it first counted a “VS Code” extension and a customer story, and four sites whose matches were not comparison pages were removed.
| What we checked | Sites | Of |
|---|---|---|
| Blog in a subfolder | 57 | 69 |
| Homepage text in the HTML | 69 | 69 |
| A price in the pricing page’s HTML | 63 | 67 |
| A page against a named rival | 55 | 68 |
| Any structured data on the homepage | 54 | 69 |
| A software or product schema node | 17 | 69 |
| Homepage served as Markdown on request | 27 | 69 |
Subfolder or subdomain
57 of the 69 sites keep the blog in a subfolder of the main site. Nine run it on a subdomain: Airtable, Beehiiv, Cloudflare, HubSpot, JetBrains, Postman, Railway, Sentry and Superhuman. Loom’s blog now lives on Atlassian’s site, and we found no blog on two. Several of the subfolders are not called “blog”: Asana’s is /inside-asana, Linear’s is /now, Ghost’s is /resources.
The agency guide the AI answer cites for its subfolder line says “search engines treat subdomains as separate websites”, and cites nothing for it. Google’s SEO Starter Guide puts the question in its list of Things we believe you shouldn’t focus on:
From a business point of view, do whatever makes sense for your business. For example, it might be easier to manage the site if it’s segmented by subdirectories, but other times it might make sense to partition topics into subdomains, depending on your site’s topic or industry.
Google publishes no “domain authority”. Most of these companies chose a subfolder and nine chose a subdomain, and Google calls it a business decision either way.
Does the page arrive without JavaScript?
Every homepage we read arrived with its text in the HTML: none carried fewer than 50 words, and the median was 1,018 words outside the navigation and footer. Pricing pages were nearly as complete. 63 of 67 carried a price in the delivered HTML. The other four carried none in the copy we were served, and on one of them, with no price anywhere in its HTML including its scripts, the page delivered five words of text: its prices load by JavaScript.
That matters less for Google than for everyone else. Google’s JavaScript documentation says it processes pages in “three main phases”, crawling, rendering and indexing, and adds: “Keep in mind that server-side or pre-rendering is still a great idea because it makes your website faster for users and crawlers, and not all bots can run JavaScript.” Vercel’s study of AI crawler traffic on its network, published in December 2024, found that “none of the major AI crawlers currently render JavaScript”, naming crawlers from OpenAI, Anthropic, Meta, ByteDance and Perplexity. Google’s Gemini, which uses Googlebot’s infrastructure, and Apple’s crawler were the exceptions.
A price that arrives only by script is a price an answer engine cannot quote. What delivered HTML means for a site in practice, and what removing JavaScript did to ours, is in What zero JavaScript did to Core Web Vitals.
Pages against named rivals
55 of the 68 sitemaps we read listed at least one page setting the product against a named rival: a “vs” page, a comparison section, or an “alternatives” page. That is a floor, since a site using another word for the section is not counted. Notion uses /compare-against/, Zendesk /why-zendesk/, Ghost /vs/. ClickUp’s sitemap lists 1,737 comparison URLs and 834 alternatives URLs.
Google holds pages like these to its guidance on reviews, which covers ranked lists and comparisons. Two of its lines are the test a comparison page has to pass: “Explain what sets something apart from its competitors”, and “Ensure there is enough useful content in your ranked lists for them to stand on their own.” A page that ranks its own product first without evidence is not what that guidance describes.
Schema, and what is missing
Structured data is the part of this advice the companies have done least. 54 of the 69 homepages carried JSON-LD in their HTML, 15 carried none, and two of those wrote it with JavaScript, where most AI crawlers will not see it. Only 17 homepages carried a node describing the product itself, a SoftwareApplication, WebApplication or Product. On the pricing pages it was 19 of 67.
Google documents a software app rich result. Its required properties are
name, offers.price, and either aggregateRating or review. For AI search,
Google’s guide to generative AI features is direct: “Structured data isn’t
required for generative AI search, and there’s no special schema.org markup
you need to add.” Schema states what a page is; it is not a route into AI
answers on its own, which is what we found when we measured it across 485
ranking pages in
Structured data for AI answers.
AI crawlers, llms.txt and Markdown on request
52 of the 69 sites take no position on AI crawlers at all: their robots.txt names none of them, so unless a blanket rule says otherwise, every crawler is allowed by default. Three block a training crawler and one blocks a crawler that fetches pages to answer live questions. A reading of the same companies’ robots.txt files on 3 September found the same picture.
46 of the 69 publish a real llms.txt file, well above the 45% we counted across 276 domains in our llms.txt study. Google’s documentation changelog says the files “won’t negatively or positively impact your visibility or rankings” in Search, and that “it’s fine if you want to maintain these files for other services or systems that use them.”
The newer practice is serving Markdown. Asked for it with an Accept: text/markdown header, 27 of the 69 homepages answered with a Markdown version
of the page instead of HTML, and 26 of the 67 pricing pages did. Asana,
Zendesk, Intercom, GitLab, Miro, Typeform, Zapier and Resend did, and
Cloudflare’s own homepage added a header with the page’s token count.
PlanetScale’s pricing page redirects any request whose Accept header lists
Markdown or plain text to /pricing.md. Cloudflare’s changelog from February
2026 describes one way it is done, at the network edge: Cloudflare will
“convert the HTML to markdown, when possible, on the fly”. We measured what the
sites serve, not which agents ask for it.
Is SEO still worth it for B2B?
“Is SEO worth it for B2B?” and “Is SEO still worth it in 2026?” both appear in the People Also Ask box on this search, next to “Is SEO dead now with AI?”. Google has answered the last one in its guide to generative AI features: “From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.”
On the evidence of these 69 sites, the companies are doing both: the ordinary SEO work, comparison pages and content in the HTML, and newer work for AI systems, an llms.txt file on 46 and Markdown on request on 27. How much the two kinds of search overlap in practice is in GEO vs SEO: the overlap, measured.
What to do, in order
- Put the blog where it is easiest to run. Google calls subfolder or subdomain a business decision. Do not move a working subdomain blog for SEO alone; nine of these 69 companies run one.
- Check the pricing page arrives with its prices. Four of the 67 we read did not. View the page source, or run the free audit.
- Build comparison pages a buyer can use. 55 of 68 sites have them; hold them to Google’s review guidance, with what sets each product apart.
- Describe the product in structured data, with a SoftwareApplication node where it fits. Seventeen homepages carried a software or product node.
- Decide on AI crawlers deliberately. 52 of 69 sites never have.
- Publish llms.txt and Markdown for other systems, if your platform makes it easy, and do not count either as Google SEO.
What this does not show
- Not an effect on rankings or citations. We measured what sites do, not whether it works. The blog count shows what large companies chose, not what ranks.
- Not all SaaS companies. The 69 are a named set of recognisable, mostly large and developer-facing companies, not a random draw. This supports “of these sites” and not a share of all SaaS.
- Not an exact count of comparison pages. The count is a floor, by URL pattern, checked by hand for false matches but not for missed ones. Five large sitemaps were read only in part.
- Not which AI agents ask for Markdown. We measured what the sites serve when asked. One site served our fetcher Markdown and a second client HTML.
- Not how Google sees the four pricing pages. Googlebot renders JavaScript. The finding is about crawlers that do not, and Vercel’s study is from December 2024.
- Not the two sites we could not read, or the editions served outside India.
Why we wrote it
The results page for “b2b saas seo” tells software companies what to do, and its AI answer opens on a press release. Measured on 69 of the companies a founder has heard of, most of that advice is already common practice: content in the HTML, pages built for buyers comparing tools, and a blog in a subfolder, the one piece Google says not to focus on. The gaps are elsewhere. Seventeen homepages describe the product in structured data, 52 sites have never taken a position on AI crawlers, and a practice no page on the search mentions, serving Markdown to agents that ask, is already on 27.