Brand Signals Graph

What is the Brand Signals graph?

The Brand Signals graph sits on your website dashboard, just below Website Performance. It puts three numbers on one chart - branded search clicks from Google Search Console, and direct traffic sessions and LLM sessions from Google Analytics - so you can see whether demand for your brand specifically is growing.

The graph plots all three series, day by day, over the last 16 months:

• Branded Clicks - clicks from Google Search queries that contain your brand keywords. Someone searching for you by name.
• Direct Sessions - visits recorded in Google Analytics 4 with no referrer: people typing your URL, using a bookmark, or clicking a link in an app that doesn't pass a referrer (messaging apps, some AI assistants, email clients).
• LLM Sessions - visits recorded in Google Analytics 4 as referrals from AI chatbots such as ChatGPT, Claude, Gemini, Perplexity, and Copilot.

Why use the Brand Signals graph?

A lot of modern discovery is hard to track directly. When someone hears about you in a podcast, a newsletter, or a group chat, that visit rarely arrives under a tidy channel - it surfaces later as a branded search or a direct visit. Traffic from AI chatbots is a mixture of the two: some of it arrives with a referrer and lands in the LLM Sessions line, and some of it doesn't, and shows up as direct.
That makes these three series useful offsetting metrics. Knowing that your brand is being mentioned is not the same as knowing whether those mentions are doing anything. If your brand is genuinely gaining ground - from AI search or anywhere else - these lines trend up. If they stay flat, that's worth knowing too.

One honest caveat: direct traffic is a noisy bucket, and no line proves why it moved. Read this chart for direction and correlation - ideally alongside your tests and annotations - not as proof that any single channel caused a change.

Before you start

The graph needs two things set up, and it will prompt you for whichever is missing:

• Brand keywords - set these in your website settings by clicking on your profile icon from the Dashboard, selecting ‘account settings’ > ‘site management’, then click the edit button next to the site you wish to add your keywords for. In the Identity section, click 'edit'. A pop up will then appear for you to check/add your brand keywords. These are the same keywords that power the Brand vs Non-Brand Report. 
• A Google Analytics 4 connection - see How to integrate Google Analytics (GA4) with SEOTesting.

Note: If only one is configured, the graph shows that series on its own until you add the others.

Reading the graph

• Use the Days / Weeks / Months buttons at the top of the Website Performance card to change the aggregation - the Brand Signals graph follows the same setting.
• Use the zoom presets (1M, 3M, 6M, 1Y, All), or drag across the chart, to focus on a period:

• Click the series buttons 'Branded Clicks', 'Direct Sessions' and 'LLM Sessions' under the chart to show or hide any of the three lines.
• Hover any point for the exact values for that day, week, or month.

Removing anomalies

Real-world traffic data contains one-day events that have nothing to do with your brand's underlying trend: a link that goes viral in a messaging group, a bot fleet crawling your site from a datacenter, or an analytics outage that records almost nothing for a day. A single day like that can be five to ten times your normal traffic and make the rest of the chart unreadable - especially in weekly or monthly view, where the spike gets baked into the totals.
When we detect days like this, a Remove anomalies button appears at the top right of the Brand Signals card. Hover it to see how many days were flagged.

What counts as an anomaly?

Anomaly detection applies to the Direct Sessions and LLM Sessions lines. Branded clicks come from Google Search Console and reflect real search behaviour, so they are always shown exactly as reported.

Each direct-traffic day is checked against the line's own recent history. A day is flagged when it is:

• more than 3x its trailing 28-day median (a one-day spike), or
• less than a quarter of that median (an outage-style dip).

Because the comparison uses a median rather than an average, normal patterns - weekend dips, seasonal growth, a genuinely improving trend - are not flagged. The check also stays quiet on low-traffic series, and on the first week of data, where there isn't enough history to judge.

What the button does

Clicking Remove anomalies replaces each flagged day with an estimate drawn from the days around it (a straight line between its nearest normal neighbours). The chart redraws instantly:

• In day view, hovering an adjusted point shows both numbers - for example, '413 (was 2,737)' - so the original value is never hidden from you.
• In week and month view, totals are recalculated from the adjusted daily data, so a single bot spike no longer inflates a whole week.

Click the button again to restore the raw data at any time. The chart always loads showing your raw data - removing anomalies is a choice you make each visit, never something applied silently.

What it doesn't do

Removing anomalies is display-only. It never changes the data stored in Search Console, Google Analytics, or SEOTesting, and it doesn't affect any other report. It's a lens, not an edit.

Frequently Asked Questions

I don't see the Remove anomalies button - why?

It only appears when the detection actually finds anomalous days in your data. No flagged days, no button.

A day I know was unusual wasn't flagged.

The thresholds are deliberately conservative so that real growth is never smoothed away. A broad multi-day surge - a successful launch week, for example - won't be flagged, because it doesn't look like a one-day glitch. That's by design: the feature targets noise, not news.

Why can I see only one of the lines?

Use the 3 series buttons under the chart to show or hide each series independently.

Why are my Direct Sessions or LLM Sessions lines missing?

Your site isn't connected to GA4 yet, or the connection needs re-authorising. See how to integrate GA4 with SEOTesting.

Do the branded clicks here match the Brand vs Non-Brand report?

Yes - both use the same brand keywords from your website settings, so the branded clicks line here matches the brand series there.

Updated on: 11/09/2026