Docs › Ama Affiliate Ultra
Every week or month the agent grades its own work: it compares this period against the last one across every connected data source, writes up what improved and what did not — then reads its own report card before planning the next round.
In one lineOff by default. Switch it on in the Progress tab and pick weekly or monthly; everything after that is automatic.
What a report contains
The comparison is always this period against the previous equal period — the last 7 days against the 7 before them, or the last 30 against the 30 before. Same length, immediately prior, so the numbers are comparable.
| Source | What it contributes |
|---|---|
| Built-in visitor analytics | Site visits, this period versus last |
| Built-in click tracking | Amazon clicks, this period versus last |
| Google Search Console | Impressions, clicks, CTR and position — when connected |
| Bing Webmaster Tools | The same, for the engine most affiliate sites forget — when connected |
On top of the numbers sits an AI-written analysis of what improved, what did not, and what the agent intends to focus on next. The Progress tab shows both: visual bars for the movement, text for the reasoning.
The report is not just for you
This is the part worth understanding, because it is what separates a report from a dashboard.
The latest report is injected into the next planning prompt. The agent literally reads its own report card before deciding what to do next. A month where updates moved the numbers and new posts did not is a month that shapes the following month’s proposals.
So switching reports on does not merely add a screen — it closes the loop between what the agent did and what it decides to do next. On a site with real traffic, that feedback is the difference between an agent that repeats itself and one that gets better.
Where it turns up
- The Progress tab — bars and full text, with previous reports kept newest-first so you can read back through them.
- Telegram — sent to your chat when the bridge is connected.
- An admin notice — which stays until you have actually viewed the report, rather than vanishing on the next page load.
Weekly or monthly
| Period | Suits |
|---|---|
| Weekly | Sites with enough traffic that seven days is a real sample, or a period of active work where you want a tight feedback loop. |
| Monthly | Newer or quieter sites. Below a few hundred visits a week, a weekly comparison is mostly noise and the AI analysis has little to work with. |
Monthly is the honest choice on a young site. A report saying visits went from 11 to 9 is not a finding, and treating it as one leads the next planning round astray.
Reading one without fooling yourself
Two habits keep these useful:
- Check the sample before the conclusion. A 40% rise on a base of 20 visits is four extra people. The bars will look dramatic either way.
- Expect lag. Content changes take weeks to move search numbers, so the report covering the week you made a change usually shows nothing. The one after it is where to look — which is also why the agent verifies outcomes on a delay rather than immediately.
Both are the same point: a report is evidence, not a verdict. It is most valuable read as a run of three or four, where a trend can actually be seen.
Common questions
Do I need Search Console and Bing connected?
No — reports work from the built-in visits and Amazon clicks alone. Connecting them adds impressions, CTR and position, which is what makes the analysis genuinely useful rather than descriptive.
Does generating a report cost AI budget?
Yes, the written analysis is an AI call and it comes out of the same hard monthly cap as everything else the agent does. It is a small, predictable cost once a week or once a month.
Are old reports kept?
Yes, stored newest-first and readable in the Progress tab. Reading several in sequence is far more informative than any single one.
What happens to reports if the licence lapses?
They stop being generated along with everything else the agent does. Reports already written stay where they are. What a lapse does is covered here.
Next
- How the agent plans its work — where the report gets used.
- Approving and rejecting tasks.
- Find the stage that is failing — the manual version of the same diagnosis.