Campaigns now target the market the client is actually in

New clients get a target market set from their first location, and partners can change it on the Locations page. Rank checks and keyword difficulty in discovery now run in that market instead of defaulting to the United States.

Most of today went to one problem: the platform assumed every client was in the United States until someone corrected it. That assumption reached further than it looked, into discovery rank checks and keyword difficulty scores, so we pulled it out at the root.

For our partners

1. Every new client is created with a target market, taken from the first location’s country, and managers can change it from the Locations page.
2. The current position column in a discovery report now checks rankings on the right Google for the market, instead of always checking the US.
3. Keyword difficulty in discovery is scored for the client’s market rather than US difficulty for everyone.
4. The AI Visibility section of the report reads better: monthly checks are grouped bars, the axis lines up with the rest of the report, and an assistant with no samples shows no bar instead of a zero.

1. Target market at creation, and a way to change it

When you add a client and pick a Google place, that place decides the market and the question never comes up. When there is no location yet, the new client form and the discovery intake show a Target market select; it disappears as soon as a place is added. If nothing says, we fall back to your own country, then to the market most of your other clients sit in.

The Locations page shows the current market and lets managers change it behind a confirm step. One thing to know: changing the market on a live campaign does not re-register keywords that are already being tracked. Those keep the market they were registered under.

Creating a client over the API accepts an optional target country as well. Spellings fold together, so England, UK and United Kingdom all land in the same place.

2. Discovery runs in the client’s market

A discovery for a UK clinic used to check its keywords on google.com with a UK town appended, which reliably came back with nothing ranked. It now checks on the market’s own Google, and difficulty scores come back for that market too. If you have run discoveries outside the US and the rankings looked empty, that is the reason, and new runs will not do it.

4. AI Visibility reads at two checks and at six

The trend was a line chart that needed hand-tuning whenever a report only had two monthly checks. It is now a grouped bar chart, one group per check, one bar per assistant, with the axis pinned 0 to 100. The questions card is titled “The questions buyers ask that we track”.

Behind the scenes

Agents got a better model of when they should wake up. A trigger used to be just a moment, and every trigger on an agent shared one set of conditions, so an agent could not both watch for an event and sweep on a date with different rules for each. Now each moment carries its own conditions and its own limits: the brief writer can hold “a plan was approved and pieces are due soon” beside “it is the middle of the month and next month is unbriefed” without the two interfering. Nothing changed behaviour on the way through; every existing setup was carried over as it stood.

Alongside that, schedules can anchor to any day of the month through the 31st, and a month that lacks the day simply skips it. The old ceiling of the 28th existed to stop end-of-month runs producing nothing useful, and the per-moment conditions say that better than a ceiling did. Calendar schedules with no hour set now land in the evening rather than mid-morning.

The other piece worth telling is Stop. If a sweep fires a batch of agent runs and we want them to stand down, that needed to be a button, not a scramble. There is now a stop on the run page and on a filtered list of runs, and a running agent checks for it before each step: it finishes its sentence, writes down where it got to, and declines every action it had queued up. Runs that had not started yet end immediately.

We also split page monitoring into two things that were wrongly one. An alert is for a page that can no longer do its job and needs a person today. A change record is for a page that is simply different from last week, which nobody needs to read until a ranking moves. Changes to content, titles, headings and schema now write to a record with a before and after and nothing to close, and the alert queue is left holding the things that actually break a page. We added two new alerts while we were there: a page that suddenly renders almost empty, and a page that gains injected spam words.