Past Maps walked away from AI money and built a 300K-user old-school web business

Past Maps overlays historical maps — from public sources like the US Geological Survey — on modern maps with adjustable opacity, built on tooling users explore for genealogy, local history, and curiosity.

In 2022 Craig Campbell, a former Meta engineer and experienced founder, sold his e-commerce tool for Shopify businesses — right as the AI boom took off. Instead of starting an AI company, he charted a website around his metal-detecting map tooling.

He exploded the public map data out in a form Google understands, tagging maps and pages so search traffic compounded. He embraced AI to run the business: a local agent model on his desktop triages customer email once an hour — cutting service time from one or two hours a day to about 10 minutes — and he is building an LLM-assisted OCR tool for historical maps, whose curving labels defeat off-the-shelf systems.

By year three Past Maps drew over 300,000 active users a month — genealogy researchers, history buffs, a daily user mapping old oil wells — and the income sustains Campbell and his wife, though he notes it matches a mid-level engineer's Meta salary, not AI-startup riches.

Organic search rewards exactly what Past Maps does — structured public-domain data made useful — and costs nothing but tagging discipline.

Subscriptions dodge both the ad-tech middleman and the AI-summary threat to page-view businesses.

The founder is his own user: a hobbyist's tool had a built-in community (Reddit's metal detecting) that seeded authentic demand.

AI on the cost side (customer service, OCR) lets a two-person operation serve 300,000 monthly users.

Subscriptions protect niche publishers from the ad-tech middleman; passion plus a useful tool plus a community that shares it remains a viable founding playbook even as AI eats the open web.

As of May 30, 2026, Past Maps sustains its founder and his wife on subscription revenue, keeps growing through Google Search, and is developing an LLM-assisted OCR tool to make more historical maps machine-readable — with Campbell insisting the human spark stays in the loop.