I’m building AdKit, an ad library for SaaS and mobile apps.
One of the problems I wanted to solve is removing friction for users. When someone adds a company or a website, I don’t want them to manually fill in forms about the brand, industry, colors, logo, or description. Nobody likes doing that.
So instead, I automate everything.
When a user submits a website, I call ScreenshotOne to capture a screenshot of the page. But I don’t only use the image — I also request metadata and parts of the HTML from the page.
From there:
- I extract CSS variables and styles to get accurate brand colors.
- I pull the favicon and assets to identify the logo.
- I pass the screenshot to AI so it can understand the company, industry, and general context of the page.
The combination works really well.
AI is great at understanding what the website is about, but it’s not always precise when it comes to exact design details like brand colors or logos. That’s where the metadata and HTML help — they give structured signals the AI can’t reliably infer on its own.
With all that data combined, I can automatically fill the brand profile for the user.
“I have my own infrastructure, so I could do it myself, but I need the screenshot at a critical time. ScreenshotOne never failed me once! And it's also much faster than my own system.”
So instead of asking them to fill a long form, the product already knows:
- what the company does
- what industry it belongs to
- its colors
- its logo
- and other branding information
That makes the onboarding experience much faster and smoother, which makes the product feel nicer and easier to use.
