Reading trends

Trend data is not a separate screen. It appears inside the pages it describes: Insights → Visibility for the visibility score over time, and Insights → Sentiment for how sentiment is moving.

Overview#

Weekly trends roll your runs up into week-over-week movement, so that what you are reading is a direction rather than a snapshot.

This exists because individual runs are noisy in a way that is easy to underestimate. AI answers vary between runs even when nothing about your brand, your site or your market has changed, so comparing one run against the next invites conclusions the underlying data cannot support. Aggregating by week smooths most of that out without hiding real movement.

Interpreting results#

Direction matters more than magnitude here. Three consecutive weeks of small gains is a substantially stronger signal than a single large jump, because a large jump is more often variation than achievement, and it has an unfortunate habit of reverting the following week.

For gaps in particular, the net number hides the story. Two weeks can move the total by exactly the same amount and describe completely different situations.

Same net movement, different weeks

Week	Gaps closed	Gaps opened	Net
Week 1	10	8	-2
Week 2	2	0	-2

Compare against your own history rather than an external benchmark. What counts as good movement depends heavily on how crowded your category is, and a number that would be excellent in a category with four players can be unremarkable in one with forty.

Expect flat stretches, too. Nothing changes between runs, and on plans with a longer cadence a given week may contain only a couple of runs in total. A flat week is usually just a quiet week rather than a failure of anything.

How much data a conclusion needs#

A useful rule of thumb: a conclusion needs three things behind it before it deserves action. Three or more weeks of readings, because two points always make a line. Enough prompts in the topic for a week to mean anything, which Prompts included this week on Visibility tells you directly. And a cause you can at least name, which is what Tracked actions and the citation record are for. Movement with all three is a finding; movement missing any of them is a note to check again next week.

The asymmetry is worth internalising: acting on a false trend costs real work aimed at a mirage, while waiting one extra week to confirm a true one costs almost nothing. When in doubt, the extra week is nearly always the better trade.

Best practices#

• A slope rather than a spike, meaning sustained direction across several weeks.
• Movement concentrated where you have actually been working. If you spent a month on one topic and that topic moved, that is about as much attribution as this kind of measurement can honestly give you.
• Gaps closing faster than they open, which is a healthier signal than the net count alone.

Next steps#

1. Look at three or more weeks before concluding anything. Two points will always make a line, whether or not there is a trend underneath them.
2. Pair movement with what you did. Tracked actions exists for exactly this, recording what you changed and when, so that a shift can be attributed rather than guessed at after the fact.
3. Investigate sharp drops before assuming the worst. A sudden fall is very often a competitor's new content being picked up rather than anything you did, and that is diagnosable through Citations.
4. Resist reacting weekly. Content and source changes take several cycles to surface, so responding to each week's movement means responding to noise with real effort.

Limitations#

Trends describe your tracked prompts on your tracked engines. They say nothing about traffic or revenue, and connecting Analytics is what gives you the downstream picture.

Weekly aggregation reduces noise without removing it. Small movements in a small prompt set remain comfortably within normal variation, and no amount of aggregation will change that.