The queue lives in a spreadsheet
Reputation work on Google reviews tends to run out of a spreadsheet, whether it’s an agency juggling client listings or an owner with two locations. Somebody pastes in a link to a review that looks wrong, and the row needs the same things every time: which business, how many stars, what date, the review text copied exactly (you may quote it later), whether the owner already replied, and whether a Google policy actually applies. After that comes a note on what to do. Often it says do nothing.
The slow part is getting from the link to the row, since each link opens Google Maps and the review is buried in a list of others. dassi opens each link in your browser, finds the review, fills in your columns, checks it against Google’s prohibited and restricted content policy, and writes back a status note or a draft owner response.
Of the review jobs we’ve seen people set up with dassi, this is the one that goes wrong least, and when it does go wrong you can see it the next morning, sitting in a cell.
What the example run found
In the illustrative run on this page, eighteen rows were marked “check” across four listings. For two of them the link opened the listing and the review wasn’t there, so dassi wrote “not visible on Sep 19” instead of “removed” (it can’t tell whether Google took the review down or the reviewer deleted it, and we’d rather not report a win nobody earned).
Five of the sixteen live reviews had a possible policy issue. Some of dassi’s status notes, as written:
| Row | Stars | Status note |
|---|---|---|
| 4 | 1 | Personal information: names a staff member and gives her mobile number. |
| 7 | 1 | No policy issue. Slow service and a rude front desk, described from a visit. Reply, don’t report. |
| 11 | 2 | Off-topic: about a local council election, not the business. |
| 15 | 1 | Possible conflict of interest: “I worked here until last month.” Someone at the business needs to confirm who wrote it. |
Row 7 is the interesting one. Slow service and a rude front desk is the review owners most want gone, and Google lets it stand because it describes a real visit, which we think is right. dassi drafted a reply for it, one of nine for reviews the owner hadn’t answered (the review response job has a recorded run of that drafting). Row 15 is harder. The reviewer could be a disgruntled ex-employee or a customer who once worked there, and nobody outside the business can tell which.
Reporting, and the pressure to report everything
Clients ask for removal, and the answer has to be that only Google decides whether a review comes down. dassi has no lever there. When a review does break a policy, a person reports it through Google’s own flow, from the review itself.
The pressure usually runs toward reporting more, and we think that’s where review management goes wrong. Mass-reporting negative reviews that describe real visits misuses Google’s report flow, and it probably makes the genuine reports from the same account easier to wave off, though Google has never said how it treats a business that reports everything. It won’t write reviews either, or draft the message offering a customer a discount for five stars. Google bans fake and incentivized reviews, and in the US the FTC’s rule on fake reviews covers both.
Running it every weekday
Saved as a workflow, the check runs each weekday morning in your browser, which has to be open then. Reviews you reported last week get a fresh dated status, and new reviews can join the queue as they arrive. Drafted replies wait in the sheet until you approve one, and then dassi posts that reply from the Business Profile. Some listings in the queue will have other gaps, like missing services or a two-line description, and we’ve written up how dassi fills those in separately.
The review text dassi reads goes to the AI model you chose, through dassi’s relay if you’re on included credits. Reviews sometimes carry a phone number or a staff member’s full name, which is often why they’re in the queue, so mention it when you tell clients about your tools. In the first week we’d read every status note before any of it reaches a client update, the row 15 kind especially, since those need someone who remembers who used to work there.