Summarize with AI
Social proof for a startup with no customers is not a row of logos. It is one specific, checkable fact: a named pilot user with a concrete result, the founders' track record, or an honest usage number with a date on it. One of those, stated plainly, does more than any strip of grey logos under a Trusted by heading.
This matters because a stranger on your homepage cannot tell your product from the ten others that launched this quarter. You know who is testing it, what changed for them and why you are the right people to build it. The visitor sees none of that unless the page says it in words he can check. Early proof is about specificity, not volume.
Four grey logos under Trusted by
Here is an example, invented for this post. A seed-stage AI tool drafts permit packets for residential solar installers. The hero reads The AI copilot for solar permitting. Below the Start free pilot button sits a thin row with the label Trusted by teams at, followed by four logos in 40 percent grey. All four are small installers. All four are on a free pilot that started in the summer. Two of them have drafted fewer than five packets.
Under the logos is a counter: 10,000+ documents processed. It counts every page of every test upload since the private beta, including the team's own. Further down, a carousel shows three testimonials signed with initials only, J.M., Operations Lead, and similar. Each says something warm about saving time. None says how much time, on what, or for whom.
Nothing on this page is false, exactly. But nothing on it can be checked either, and a visitor who sells solar for a living will notice that within one scroll.
What a stranger reads in that row
The founder reads the strip as four companies that chose him. A stranger reads it as four names he has never heard of. Unknown logos do not borrow anyone's reputation, because there is no reputation to borrow. They only tell the visitor how big your customers are, and the honest answer is small.
The heading makes it worse. Trusted by is a claim about a relationship. A free pilot is a trial, and the visitor knows the difference the moment he asks himself whether these teams pay. The counter fails the same test. Ten thousand documents compared to what, over what period, by whom? A number with no baseline reads as decoration.
Most advice on early social proof says to get as many logos and testimonials as you can, as fast as you can. That advice is wrong at this stage. Volume is the one thing you cannot have yet, so faking it puts your weakest point in the most visible place on the page.
Why the strip stays up anyway
Watch what the founder does. An advisor told him the homepage needed social proof, so he messaged his pilot users on a Friday asking if he could use their logos. Three said yes. The fourth did not reply, and he added it anyway because the row looked thin with three. On sales calls the first question is often who else is using this, and he answers by naming the pilots out loud, with details the page never mentions.
He has the proof. It lives in his head and in his call notes, and the page shows a weaker version of it. This is the same gap that runs through every early website: he knows his product and the evidence for it, and nobody else does until it is written down where a stranger can see it.
One named pilot with one number
The rewrite drops the row and replaces it with a single block. In the same example, with a name made up like the rest of it, it reads like this. Kestrel Solar, an 11 person installer in Tucson, used the tool on its first 60 jobs this summer. Packet prep went from about three hours per job to under an hour. Below that sits the name and role of the person who agreed to say it, and a link to a short write-up of how they used it.
That block is smaller than the logo strip and it does more. It names a real kind of customer the visitor can picture. It gives a result in units he already counts, hours per job. It is narrow enough to be believed, and it tells the right reader that this is for an installer about his size. The wrong reader, a utility with a procurement team, learns that too, and that is useful.
Ask for permission the right way. Do not ask a pilot user if you can use their logo. Ask if you can say what changed for them, in one sentence, with their name on it. Send the exact sentence you want to publish. People say yes to a specific line far more easily than to an open request, and the line is what you need anyway.
Proof you have before anyone pays
If no pilot has a result yet, you still have three kinds of proof. Use the ones that are true and skip the rest.
The founders. Who built this, and what did they do before? A line like two founders who ran permitting at a 200 person installer for six years is proof of understanding, and for a product with no users it is often the strongest thing on the page. Our site for Mi-VAD, a heart pump company that cannot point to patients yet, leads with exactly this: the surgeons and engineers from Mayo Clinic and Cornell behind the device, and the patents granted on it. The page speaks to investors and clinicians at once without claiming a single customer.
Honest usage. A small number with a date and a definition beats a big one with neither. 212 permit packets drafted by pilot teams in September, updated every Monday, is a number a visitor can believe. If you change the definition later, change the label.
Visible progress. A public changelog with dated entries shows the product is alive and moving. It costs nothing to keep, and a visitor who sees shipped in the last week reads it as momentum without you having to claim any.
What weak proof costs before a raise
Every visitor to the homepage was paid for, in ads, in content or in hours of posting. A proof section that cannot be checked turns some of those visitors back at the exact moment they were deciding whether to trust you. The cost lands whether or not anyone tells you.
It also shows up in the raise. The seed bar has risen to roughly 300 to 500 thousand in ARR for SaaS and product-led companies, and across all sectors the median gap from seed to Series A is around 616 days. Roughly 70 percent of seed companies never raise a Series A at all. An investor who clicks from your deck to your site is checking whether the story holds. Four free pilots dressed up as customers is the kind of detail that turns into a question on the next call. One named pilot with a real result is the kind that answers one. If you want the wider picture of what reads as a serious company, how to make your AI startup look like a YC company covers the rest of the page.
The rewrite, in order, this week
- List every claim in your current proof section. Next to each, write how a stranger could check it. Cut any line where the answer is that he could not.
- Pick the one pilot user with the clearest before and after. Write the sentence you want to publish, in their units, and send it to them for a yes.
- Write one line about the founders that a stranger could look up. Previous company, role, years. No adjectives.
- Replace any counter with a number that has a definition and a date. If the honest number is small, show it anyway or remove the counter.
- Change the heading. Trusted by becomes something accurate, such as Piloting with or In use at, until those teams pay.
- If you are pre-launch, put the proof next to the signup, not below the fold. A waitlist page for an AI startup lives or dies on whether the visitor believes the team can ship.
And never fake it. No stock photo headshots on testimonials, no logos of companies that only took a demo, no quotes you wrote and someone approved without reading. A founder in your category will spot one fake in seconds, and after that he discounts everything else on the page.
If you would rather compare who could do this rewrite for you, the landing page agencies for pre-seed AI startups ranking is a fair place to start.
Studio Maydit designs websites for AI founders in the US, UK and Europe, and most of that work is helping a stranger understand a company the founder already understands completely. That includes turning the proof you already have, a pilot, a background, a number, into lines a visitor can check in the first scroll. We work in fixed scopes of three to four weeks, in Framer, Webflow or custom code. If your proof section is a row of logos you would not want a prospect to ask about, book a 30 minute call with Studio Maydit.
