What 100 startup homepages are missing

At the start of October my studio scored 100 funded AI and SaaS homepages against the twelve questions we use with clients. Sixty were seed-stage companies from Y Combinator’s 2026 batches and forty were Series A or later from the a16z portfolio. We captured every site in a browser at desktop and phone width on 2 October, gave each question 0, 1 or 2 points, and wrote the rubric before we looked at a single homepage.

The average came out at 14.1 out of 24, and nobody reached 20. Later-stage companies did better, 15.9 against 12.8 at seed, which you’d expect from teams with more time and money. I was more interested in where the points went missing.

They look the part

Most of these sites look expensive. 81% got full marks for looking as established as their funding, and 76% looked like the same company across their site, product and socials. If you’ve raised a round you can usually afford a good-looking homepage, and it shows.

They hide the evidence

Proof was the weakest question in the study by a distance. 53% of the homepages showed no customer logos, testimonials or case studies at all, and only 6% got full marks for it. At seed it’s starker: 58% put an investor badge in the first screen, and 23% show any customer proof there.

I understand why. A fund’s logo is often the most impressive thing a young company has. But someone reading your homepage is trying to work out whether this works for a company like theirs, and an investor logo doesn’t answer that. A named customer with a number does, and only 21% of the sites paired the two.

The phone version is worse

On desktop, 35% of sites passed the five-second test, meaning a stranger could tell what they do almost at once. On a phone that fell to 21%, and 15% of sites scored lower on mobile overall. Headlines written for a wide screen wrap into something vaguer at 390 pixels, and the product shot tends to slide below the fold.

What I’d fix first

If you change four things on your homepage this quarter, I’d make it these.

Put proof in the first screen. A named customer with a result does more than an investor logo, and most seed sites have that backwards.

Write the headline on a phone first. If it says what you do at 390 pixels wide, it’ll say it everywhere.

Name the buyer and one real difference above the fold: a role or an industry, plus a number, an alternative or a mechanism that sets you apart. Only 26% of the sites named a specific buyer in the first screen, and 22% stated a difference.

Show the product. 19% of the homepages had no product UI anywhere on the page. A real screen or a short video beats an abstract illustration.

What the study can’t tell you

It looks at homepages only, captured once, on one day, over a fast connection. A model scored six of the twelve questions, and against my hand checks it scored low more often than high, so treat the clarity and proof figures as conservative. The sample is random within two investor portfolios, which isn’t the same thing as every funded startup, and no company is named next to its score. The method and every chart are on the research page.