2
10 Comments

I audited 50 B2B sites for AI visibility. The best-designed sites were the least cited. Here's what I learned.

Hi Indie Hackers,

For the past 2 months I've been auditing whether ChatGPT and Perplexity actually cite B2B companies when buyers search for their category.

I've scanned ~50 SaaS and marketing sites so far.

The result completely broke my mental model.

The best-designed websites were often the LEAST cited by AI.

One site with a polished 92/100 design score appeared in only 8% of AI answers.

Another site with a messy 34/100 design score appeared in 47% of answers.

At first I thought the issue was technical SEO.

It wasn't.

I started noticing there are actually TWO separate layers behind AI visibility:

LAYER 1 — Can AI access and extract your content?

- AI crawler access

- Schema

- Content structure

- Information density

LAYER 2 — Does the internet validate your brand?

- Reddit discussions

- G2 reviews

- Product Hunt launches

- Wikidata presence

- Press mentions

The "ugly" site dominated Layer 2.

People talked about them everywhere.

The polished site failed Layer 2 almost completely.

Nobody outside their own website was validating them.

That was the moment the pattern clicked:

AI systems don't just rank websites.

They infer trust from the entire internet.

The homepage matters less than I expected.

External validation matters far more.

Curious if others building in SEO / AI search / SaaS are seeing the same thing.

Drop your domain below and I'll run a quick AI visibility check on it.

Jean EA, founder of Zypact

zypact.com

  1. 1
    This matches what I have been seeing too: design quality and AI-readability are not the same thing. A site can look strong to humans but still make AI assistants piece together the offer from scattered service pages, pricing sections, FAQs, policies, and CTAs. I think the useful first step is a simple diagnostic: can ChatGPT, Gemini, Claude, or Perplexity clearly understand what the business offers, who it serves, what the next action is, and which pages matter? That is the angle we are working on at AIUNSEEN Studio. Not ranking guarantees, just making public website information easier for AI systems to read and summarize.
  2. 1

    I think the broader shift is that AI visibility is becoming a reputation problem more than a website problem.

    A website can explain who you are, but the rest of the internet determines whether that explanation is credible. The companies that understand that distinction will probably have a much bigger advantage than those focused only on optimizing pages.

    1. 1

      Exactly this.Websites explain what you do.The rest of the internet decides if it's true. That's why brands with mediocre websites but strong external presence dominate AI results.

      The companies that treat AI visibility as a reputation problem, not an SEO problem, will win the next 5 years of buyer acquisition.

      what industry are you in?

      Would love to see if this pattern holds in your space.

      1. 1

        I'm building for founders and B2B SaaS teams.

        Across the companies I've looked at, the same pattern keeps showing up: AI seems to trust brands that have consistent third-party validation more than brands with polished websites alone.

        That's what made me start thinking of AI visibility as a reputation problem rather than an SEO problem.

        Have your audits shown any industries where that pattern breaks down?

        1. 1

          Two industries where the pattern partially breaks down: Highly technical niches (dev tools, security, infrastructure) Layer 1 matters more here. When I've looked at how models cite infrastructure tools, technical depth in documentation pulls weight that Reddit chatter can't easily replace. A well-structured docs site can outrank brands with louder community presence not always, but enough to matter.

          Newer AI-native categories: When the category is still forming, third-party coverage hasn't accumulated yet. Layer 1 becomes the default signal by default not because it's trusted more, but because there's nothing else to read.

          For established B2B categories SEO, marketing, HR, sales Layer 2 almost always dominates. Models pull from what other people have said about you, not from your own positioning. Your About page is the last thing that moves the needle. What kind of B2B SaaS founders are you building for?

          1. 1

            Mostly early-stage B2B SaaS founders.

            Your reply actually made me think about a consequence of those exceptions that I hadn't considered before. I don't think it's really an industry distinction—it's something else that's shaping how companies should think about AI visibility as the market evolves.

            I don't think I can explain the reasoning properly in a thread without flattening it.

            If you're open to it, what's the best email to reach you on?

            1. 1

              This comment was deleted 2 months ago

              1. 1

                Thanks. I don't actually use WhatsApp much.

                If you're okay continuing over email instead, what's the best email to reach you on? I think it'll be much easier to explain the reasoning properly there.

                1. 1

                  contact@zypact.c

                  1. 1

                    Thanks! I’ve just sent it over.

                    Looking forward to hearing your thoughts whenever you have a chance.