Hear it from your companion, not from a customer

What is the exact prompt?

Written as a routine from day one. The schedule is part of the prompt:

• "Every weekday at 7:30, check these sources for mentions of [company], [product], [CEO name], and the misspelling [variant]: Google News, Hacker News, Reddit search, [two trade publications], and X search in my logged-in session."
• "Include only items published in the last 24 hours. For each: headline, source, one sentence on why it matters to us, tone (positive / neutral / negative), and the link. Order by consequence, not recency: a customer thread with traction outranks a syndicated press mention."
• "If there is nothing, say exactly: no mentions in the last 24 hours. Do not pad the report with old news or near-misses. Never reply, comment, or vote anywhere: read only."

Why the report says "no mentions today"

A report that always finds something is a report you stop reading.

Say the words and a quiet week comes back as one line, which is what keeps the noisy week worth opening.

The read-only line does the same job for scope. The sweep watches and you decide when to answer. Consequential actions pause for approval anyway, and Ask Always in Settings → AI permissions widens that gate.

A good week is mostly empty

A truthful sweep for a small company reports nothing most days, catches the Tuesday podcast mention you would have found on Friday, and flags the negative review while the thread is still answerable. The value sits in maybe four reports a month, which is exactly why a person quits and a routine does not.

When you learn about a mention some other way, check whether the routine caught it. The fix is usually one more source or name variant. Teams watching rivals run competitor monitoring on the same schedule, and the community side is brand-mention tracking.

What should you check on the tone labels?

Click through every "negative" for the first two weeks.

Sarcasm and faint praise are where tone classification is weakest, and a mislabeled negative costs you either a false alarm at 7:45am or a real fire filed as neutral. The why-it-matters line is your fast check: if the sentence does not justify the label, read the source.

Once a month, search one source by hand and compare against that day's report. Monitoring fails silently when a login expires or a site moves its search. Wider market context belongs to the market research playbook.

Frequently asked questions