I could not see my own landing page with fresh eyes anymore, so I built five strangers to do it for me
After you have looked at your own landing page twenty times, you cannot see it anymore. You know where the pricing link is, so it feels obvious. You know what the product does, so the headline reads fine. Every founder I know has this blind spot and none of…
After you have looked at your own landing page twenty times, you cannot see it anymore. You know where the pricing link is, so it feels obvious. You know what the product does, so the headline reads fine. Every founder I know has this blind spot and none of us can fix it alone. So I built FiveStrangers. You give it a URL and pick what visitors are supposed to do there (sign up, call you, buy something). Then five AI visitors open your site in a real browser, one at a time, each with a different personality: - the Google skimmer, who gives your page ten seconds before leaving - the mobile ad clicker, browsing on a phone with no patience - the skeptical evaluator, who reads everything and hunts for proof - the comparison shopper, who has three competitor tabs open - the ready buyer, who just wants to convert and counts every obstacle They scroll, click your buttons, follow links across pages, and narrate what they are thinking the whole time. At the end each one either converted or bounced, and you can see the exact moment you lost them. While that runs, the boring measurable stuff happens too: Lighthouse performance, accessibility, broken links, SEO, security headers, and whether ChatGPT and Perplexity are even allowed to crawl your site. All of it turns into one score with the evidence attached. Two things it found on a real site I tested it against, a nursing service in Berlin: The skeptical visitor read the entire site, then refused to call. Its reasoning: the page says "certified nursing service" but never names a certificate, an authority, or a date. For a business where families are handing over care of a parent, that gap killed the conversion by itself. The comparison shopper clicked the "call us now" button on the contact page and nothing happened. It focused a form field instead of starting a call. Nobody had noticed. That is a real bug that was quietly costing phone calls. What it is not: a guarantee, or a replacement for talking to actual users. The findings are automated, and I make it cite what a visitor actually saw or a number that was actually measured, because "your CTA could be stronger" helps nobody. Status: my payment provider has not approved the store yet, so it is early access at the moment. You can put in your site and I will email you the moment audits open. There is also a complete example report you can download without giving me anything, so you can judge the output before deciding. >>> fivestrangers.io Price when it opens is 12.99 one time per audit, no subscription. Happy to answer anything about how the agents work, and honestly interested in hearing where you think this falls apart.
Collected discussion
Exactly! This audit helps you to fix these kind of things, before you share your new website. ☺️
Oh, and your site needs a favicon.
Damn. This looks nice tho. Funny as well. Where can I check this partner integration. Did not find it 😅
Thanks! Am still improving the agents, to get a better result.
keeping building , bes of luck mate!
I like the concept! Particularly from the marketing side. I do think the 5 strangers might be inherently challenging against a competitor though who say, does ten strangers or a landing page audit tool that scans thousands.
Thousand agents are just too much. Would increasd the costs a lot.
Great Idea! Will look into it :) Thanks!
Keep confirmed defects and simulated reactions in separate sections of the report. A broken call link is a defect. The claim about an unnamed certification is a hypothesis to test with people in that buyer group. Let the score weight confirmed failures, and show hypotheses with the page evidence that produced them. That keeps buyers from fixing generated opinions before real conversion blockers.
Have you seen uglybaby.io? I've used it a few times and it's neat. They might have a partner integration you could integrate in as well.
That moment when you stare at your own page so long the call button is broken and you dont even see it, i felt that in my stomach the skeptic refusing to convert because you never named the certificate is exactly the kind of thing that eats conversions but nobody in-house catches. i used to work in marketing and we had signup forms with dead fields for weeks until a customer emailed us mad
this is an amazing concept. Founder blind spots are real, and simulating distinct user personas is a clever way to bridge the gap between static design audits and actual user journey testing.
Good instinct already in this thread to separate confirmed defects from generated hypotheses, that's the right split. One thing that makes the hypothesis side less generated-feeling over time: instead of a one-off validation round, keep a standing reference of what your actual buyer group says unprompted, in reviews of you and competitors, forum threads, comparison posts, and recalibrate the personas against that periodically instead of just their training priors. I've been doing something similar with Hugo on the customer research side, the gap between what a simulated visitor assumes and what real buyers actually say tends to drift over months, it doesn't stay fixed after one round of testing. Doesn't replace what you built, just keeps the simulation grounded instead of confidently wrong six months from now.
that skeptical evaluator's tough. we swapped our stock photos for raw, unedited group shots of our own team and conversions went up. people can smell a template from a mile away.