A monthly ten-minute routine that turns public competitor information into a one-page brief: what they changed, what it signals, and the one move worth considering. Prompt included.

You don't need surveillance. You need a routine.

Most solo founders relate to competitors in one of two ways: total avoidance ("I just focus on my own work" — noble, occasionally expensive) or doom-scrolling a rival's Instagram at 11pm (research theatre; produces anxiety, not decisions). Both miss the same point: competitor awareness is a system job, not a mood.

The decoded idea: the value isn't the data, it's the delta. A competitor's website today tells you little. Their website compared with last month tells you plenty: a price quietly raised, a service dropped, a new niche in the headline, three job ads for delivery roles. Changes are strategy leaking in public. And tracking change-over-time is precisely the boring, structured work that belongs in your AI system — the analyst role, on a monthly schedule.

The routine — ten minutes, once a month:

1. Pick five competitors, once. Three direct, one aspirational (where you want to be in two years), one adjacent (serves your customer a different way — often where disruption actually comes from). Save the list; don't relitigate it monthly.
2. Collect public pages only. Homepage, pricing page, services page. Copy the visible text. Public information, gathered the way any customer could — that's the whole scope, and it's plenty. No scraping tools, no fake enquiries, nothing you wouldn't be comfortable explaining.
3. Run the scan prompt (paste this month's text plus last month's saved version):

"Compare these two snapshots of the same competitors' public pages, taken a month apart. For each competitor: (1) what changed — pricing, offers, positioning language, anything added or removed; (2) the most plausible boring explanation, then the most plausible strategic one; (3) rate how much this matters to a [your business] serving [your customer]: ignore / watch / respond. Finish with one page maximum: the single most important shift this month and one move I could test in response. If nothing meaningful changed, say exactly that — do not manufacture insight."

1. File the snapshot. Next month's scan needs this month's baseline. The archive is the asset — six months in, you can see trajectories, not events.

That last prompt line matters more than it looks. Models are eager to please, and an eager analyst invents patterns. Telling it that "nothing changed" is an acceptable answer is the difference between intelligence and horoscope.

What to do with the output. One page, one decision: ignore, watch or respond. Most months: ignore. That's success, not failure — you bought certainty that nothing needs your attention, for ten minutes. The occasional "respond" month is where the routine pays for a year of itself: you see the price move or the niche shift while it's a signal, not once it's a trend everyone's chasing.

The responsible-AI note. Public information only — the rule stated plainly in the routine is also stated plainly here. Verify anything surprising against the source before acting on it (the model can misread a snapshot). And keep your own strategy notes out of the prompt beyond your one-line context; the analysis needs your market, not your secrets.

Reading signals without paranoia. A few translations from six months of running this routine. Competitor drops their prices → check whether their delivery promise shrank too; a price cut with a scope cut is repositioning, not aggression. Competitor raises prices → the most useful signal on the list; someone has tested your market's ceiling for you, free. New niche language on their homepage ("for dental practices", "for trades") → they've found focus pays; the question is whether they've vacated ground you want. Hiring pages beat press releases every time: job ads are money committed, announcements are money hoped for. And a competitor suddenly posting AI-generated content at volume tells you about their capacity, not their strategy — volume without a point of view is the easiest thing in the world to out-position now.

The trap to name. Awareness curdles into imitation quietly: three months of scans and you start pricing like them, writing like them, chasing their niches. The routine's job is the opposite — to know the landscape precisely so your different bets are chosen, not accidental. One guardrail: every "respond" decision must name the customer it wins, not the competitor it answers. If you can't say who it's for, it's reaction, not strategy — file it under watch and move on.

One decoded AI move, every week. Follow @aithebrief on Instagram for the daily version, or subscribe free on Substack for the deep one.