Making AI detection readable

GPTZero was a consulting project. Through 2024 and 2025 I worked with Faizan Mehdi, and with co-founder Ed Tian. I was not on staff. GPTZero is an AI detection platform. It helps people see whether a piece of writing was generated, and it sits in the middle of a loud public argument about that.

The scope was the homepage and the graphics around it: social posts and blog imagery. Not the model, and not every screen in the product.

The homepage had to say what the product is for before it asked anyone to paste a paragraph. Hierarchy, type, and the first picture of the product did that work. Social and editorial then had to look like the same company, not a separate campaign.

These are the public homepage during the engagement, not reconstructions. July keeps the scan beside the claim. November leads with the line the company still uses: more than an AI detector, preserve what’s human. A simple checkerboard stands in for the idea.

Further down, the page teaches the difference with two sentences, one human and one generated, then shows what a deeper scan actually inspects. The product claim stays on the page. I am not treating those accuracy figures as a result of the design.

Social creative was the second workstream: campaign graphics, educational posts, and cuts for each channel. The point was to reuse the site’s type, color, and the contrast between human and generated writing, rather than invent a second brand for the feed.

Those frames live in the shared Figma file with Faizan. I am not filling this section with stand-ins.

The third workstream was blog and editorial graphics: headers, featured images, and smaller compositions for pieces that explain detection, writing, and the product. Faizan was also publishing on GPTZero’s news site through this period, so the visual work sat next to that writing.

A header had to make a technical note feel approachable without turning it into a cartoon. Same rule as the homepage: the idea first, the lab equipment second.

Full disclosure: this was consulting. I was not a GPTZero employee, I did not lead the company, and I did not own every deliverable. Faizan Mehdi, marketing strategy consultant, was the partner on the work. Ed Tian, co-founder, was the stakeholder we were designing with.

The constraint was the subject. Detection is a technical claim, and a teacher or a writer still had to understand the page immediately. As a consultant, working beside Faizan and with Ed in the conversation, the engagement stayed in its actual shape: a homepage, a social system, and editorial graphics, made together.