Skip to main content
Anymorph has several numbers that look alike. This page is a cheat sheet for picking the right one for the question you actually have.

The five numbers

The first three tell you how AI talks about you. Citations tells you what AI read to say it. Performance tells you whether anyone actually showed up.

Which number for which question

Don’t get the direction backwards

Mention rank is the only one where lower is better. Moving from #2.4 to #1.8 is an improvement. The screen flips the colour of that metric’s change badge for exactly this reason, so green always means “better”. Everything else: higher is better.
Your Mention rate can be good while your Mention rank is poor, and vice versa. That’s the difference between being named often but always last, and being named rarely but first when it happens. Read the two together.

The period-comparison trap

There are two different kinds of change figure in the product, and they sit close enough together to confuse people. That topic column isn’t “this topic versus last week”. It’s “is this topic trending up or down inside this window”. The two figures moving differently is not a bug — they answer different questions. One more thing: if the previous window had no measurement at all, the change is left blank rather than shown as 0. A means “couldn’t be read”, not “flat”.

Change the prompts and every number moves

Every metric is calculated against the questions you track. Change the questions and the denominator changes, so the numbers move even when nothing changed for the brand.
Don’t read across a prompt panel version change as one trend line. A new version asks different questions. It isn’t the same measurement. Compare within a version.
Two ways to check:
  • On Overview, read the Measurement basis strip. If Prompt changes is not zero, or there’s a v[n] published badge, the question set changed inside the window you’re looking at.
  • On Visibility, hover the dashed vertical line on the chart. The tooltip names the change, e.g. Prompt panel version changed.
Region and topic filters change the denominator the same way. When you compare, keep the filters identical.

A 15-minute weekly review

1

Skim the week on Overview (3 min)

Read This week’s drafts for what got written and how many drafts are waiting on you. Then check Prompt changes in the Measurement basis strip. If the question set changed this week, this week’s movement isn’t a like-for-like comparison.
2

Find the losing topics on Visibility (5 min)

Open Visibility with the period on 7d. Read the big number and the Share of Voice (SoV) bar. Then in Mention rate by topic, click Gaps for the five topics where competitors lead you most.
3

Find your next placement on Citation (4 min)

On Citation, see which source type AI cites most. Filter the table to that type and work down by citation count — that’s your shortlist of where to be next.
4

Clear the review queue (3 min)

Back on Overview, click On-site (n) or Off-site (n) and review the drafts waiting for you.
Still asking “but why?” — open Visibility’s lightbulb Insights panel and click a card’s Mentions link. Reading the actual AI answers is the fastest way to find out.

The monthly routine

1

Run the same review on 30d

Same order as the weekly one, but set the period to 30d. The daily bounce washes out and the real trend shows.
2

Pull the report

On Report, click Download PDF for a document you can share. Hit Regenerate first if you want it built on the latest analysis. Each workspace keeps one report — the latest.
3

Check what it returned on Performance

On Performance, look at AI referral traffic and whether the pages you published are actually getting cited. The Impact tab shows whether mentions rose after a page went live.

Things people commonly mix up

Mention rate is a count: what percentage of answers named you. SoV is a share: how much of the brand conversation was yours. SoV grows when you’re named early and discussed at length. If you’re always mentioned but only in one line at the very end, your rate is high and your SoV is low.
No. A mention is your brand name appearing in the answer text. A citation is a link the answer leaned on. An answer can cite your page without ever naming you. They count different things and will never match.
The one in the Mentions detail panel is your share inside that single answer. The one on Visibility is your share across every answer in the selected period. Same name, different numbers.
Their default periods differ — Overview starts at 30 days, Visibility at 7. The region, topic and engine filters are also set per screen. Match the filters before you compare.
It isn’t tracked. Untracked brands still count in the Mention rank pool, so they can push your rank down, but they don’t appear in SoV or on the scoreboard. Add it as a competitor in settings to bring it into the comparison.
It means the figure couldn’t be read for that scope. It never means zero. Usually there was no measurement in the previous window, or that card’s data source isn’t connected, or the metric simply can’t be measured per topic.
No. A prompt that names your brand is weighted at 10% of a normal one, and the topic breakdown excludes branded and comparison prompts entirely. To move the numbers you have to win the questions where nobody named you first.

Visibility

Mention rate, mention rank and share of voice.

Citation

The sources AI answers pull from.

Mentions

Read the AI answers themselves.

Performance

Traffic and results from published content.