TYPESCOUT

Font libraries organize typography around characteristics like serif, sans-serif, display, and weight. Those classifications are useful, but they’re not necessarily how a designer thinks when building a brand.

I wanted to ask different questions: What are finance brands using when they want to communicate trust? What typefaces repeatedly appear across technology companies? Is the font I’m considering actually unusual—or is everyone already using it?

Answering those questions required connecting typography to the brands already using it.

Old model

Font

Imagine where it might work

TypeScout

Brands

Observe what they’re actually using

A font specimen can show how a typeface looks. A brand profile shows how it behaves as part of an identity.

TypeScout captures the visual context surrounding each brand and connects it to the typography detected on the site. Industry and vibe classifications provide another layer of context, while separate heading and body analysis reveals how different typefaces are being combined within the same system.

Discovery also works in reverse. Selecting a typeface opens its profile and reveals every brand currently using it, turning an individual font into a collection of real-world references.

Instead of evaluating Inter as an isolated specimen, for example, a designer can see how it appears across technology, finance, legal services, energy, and other industries—and whether its ubiquity makes it the right choice for the project at hand.

Filtering by industry and vibe reveals the typefaces appearing within a specific branding context. Rather than assuming what “finance” or “trustworthy” typography should look like, designers can inspect the choices brands in that intersection are already making.

The result isn’t a prescription. It’s a reference point.

TypeScout also counts how many brands in the directory currently use each typeface. That turns popularity into another design input.

A heavily adopted font may provide familiarity, accessibility, and an established visual convention. A rarely used typeface may create distinction—or simply be uncommon for a reason. TypeScout doesn’t decide which is better; it makes the tradeoff visible.

As the directory grew, I deliberately kept classification mutually exclusive. Every brand receives one primary industry and one primary vibe rather than appearing across every category that could plausibly apply.

The constraint sacrifices some nuance, but protects the usefulness of the dataset. If the same brand repeatedly appeared across overlapping classifications, category counts would become inflated and patterns harder to trust.

At scale, consistency became more valuable than exhaustive labeling.

TypeScout currently shows how typography is being used across its growing directory. Because the system continues discovering and recording brands over time, that same structure creates the foundation for something more interesting: observing how those patterns change.

A font that dominates an industry today may decline. Another may quietly spread across categories before becoming an obvious design trend. The longer the directory operates, the more opportunity there is to study those movements historically.

01 — Discover

Find potential brands across targeted industries.

The first challenge was extraction. Website code is inconsistent, typography isn’t always declared cleanly, custom fonts appear under unpredictable names, and visual content can change depending on loading behavior, cookie banners, or bot protection.

Rather than relying only on CSS declarations or metadata, TypeScout renders each site and inspects the computed typography being displayed to the user. Detected fonts are then resolved against known libraries before the brand moves deeper into the pipeline.

As the directory scaled, I ordered the pipeline so inexpensive decisions happen before expensive ones. Previously processed domains and regional duplicates are rejected before rendering or AI evaluation, preventing unnecessary scraping, model calls, and repeated work.

A persistent processed-domain record also means TypeScout remembers failures and rejections—not just successful brands—so the same bad candidate doesn’t consume resources every time discovery encounters it.

Early autonomous discovery immediately exposed a bias: technology companies dominated the results because they’re highly visible online and easy for search systems to find.

That would have made TypeScout larger without necessarily making it better. I wanted the directory to reflect typography across nonprofits, beauty brands, finance, retail, healthcare, and other categories—not simply whichever companies were easiest to discover.

I restructured discovery around category coverage, allowing underrepresented industries to receive priority as the directory grows.

Website screenshots remain the least reliable part of the pipeline. Some sites load slowly, block automated browsers, present cookie or CAPTCHA layers, or delay visual content beyond the capture window. I improved timing and failure detection, but the problem isn’t completely solved.

Rather than hiding that limitation, TypeScout treats imperfect extraction as something to continue auditing as the dataset grows.

5 Days

From idea → working product

The result is a system I no longer need to operate manually. Discovery, extraction, classification, and publishing continue in the background while the directory grows on its own.

2,951

Brands Tracked

1,474

Fonts Cataloged

15–40

New Brands / Day

20

Discovery Categories

0

Manual Publishing Required

TypeScout has only recently begun appearing in search results, but early organic visitors are already using the directory across the behaviors it was designed around—browsing brands, sorting results, interacting with fonts, and filtering the dataset.

At this stage, I’m treating that activity as product observation rather than traction. The more important signal is whether people who arrive understand what the tool is for and actually use the research interactions available to them.

48%

30-Day Bounce Rate

Observed interactions

Typography trends rarely announce themselves. They accumulate through thousands of individual design decisions until a typeface that once felt distinctive suddenly feels ubiquitous.

I remember when Gotham seemed to be everywhere. Years later, fonts like Montserrat became increasingly common across digital brands. Those transitions are easy to recognize in hindsight, but much harder to observe while they’re happening.

Because TypeScout continuously records typography alongside brand and industry context, the dataset creates the foundation for studying those changes over time.

TypeScout doesn’t have enough historical data to make those claims yet. The opportunity is in letting the record accumulate.

Typography adoption over time

Conceptual — illustrating future longitudinal analysis

Fonts sit unusually close to the foundation of a brand identity. Tracking their adoption won’t explain every shift in design, but it can provide one measurable signal of how visual conventions move across industries.

At sufficient scale and over enough time, TypeScout could become less about discovering individual fonts and more about observing the typography layer of the web itself.

I started TypeScout because I wanted to build something genuinely useful for creative people—and to test whether I could take a problem beyond creative direction into product design, information architecture, and working infrastructure.

Five after-work days later, the product was live. Since then, the most satisfying part has been watching it continue doing the job without me.

Building TypeScout reinforced something I’ve increasingly brought into my creative practice: I don’t have to stop at defining the idea or directing its execution. I can design the system that makes the idea work.