Overview

This is also where you land when you open a project.

Overview#

Overview is the summary layer, and it exists to answer one question in about ten seconds: has anything changed since the last time I looked?

It is deliberately not a place to do work. Every number here appears in fuller form on the screen that owns it, and the habit worth forming early is treating this page as a reason to open one of those screens rather than as an answer in its own right.

Panels and controls#

The KPI row across the top is four cards. Each shows the latest week with runs, the 7 days ending at your most recent completed run and the same week as the Sentiment and Visibility screens, with its change against the previous week. The week's dates sit above the cards.

The KPI row

Card	The question it answers
GEO Score	How discoverable are we to AI engines, overall?
AI Discovery Reach	How often do we actually surface?
New Brand Citations	What started citing us in the latest week?
Brand Sentiment	How are we being described?

Below the row, the visibility chart is paired with Competitors, which shows your tracked competitors ranked by Share of voice (latest week): who else the engines name is part of reading your own trend line, not a separate errand. Each row gives the company's name with its domain under it, and your brand goes by its name in Brand Hub. The figure covers the latest week with finished analysis, with its change in points against the previous week, and it is the same figure the Tracked competitors page shows for that week. SOV rank is the company's place by share of voice among every company named in AI answers that week, not only the ones you track. Under those, a trends panel tracks direction and two widgets surface work waiting for a decision, Suggested content from the Content Scout and Brand reputation actions from the sentiment side.

The last row pairs LLM Performance Index, which breaks performance out per engine with a Mention Rate and a Sentiment Score for each, both for the latest week with the change against the previous week and an eight-week line beneath, so a healthy average hiding one weak engine becomes visible, with Top citation sources. The index lists only the engines that ran in the latest week. That lists the five sites the engines lean on most when they answer about you, in three columns: the source, its Influence as a banded bar with the score beneath it and an info icon explaining what the score weighs, and its Reach, meaning how many AI answers in your category cited that source in the latest week, with a small arrow and number when that changed against the previous week and a badge saying how many of those answers cited you. Influence is measured across all your runs, so it holds steady while Reach moves week to week. Those are different numbers on purpose: a source can shape a dozen answers about your category and have mentioned you in none of them. A source that sits outside that measurement, or was in none of the latest week's answers, shows how many of its pages were cited instead, which is a different question and is labelled as one, rather than repeating the badge's number. On a wide card a fourth column, Favoured by, names the engine that leans on each source most, adjusted for how many prompts each engine answered. On a narrower card the same engine sits on its own line under the source instead. A dash, or no line, means no engine could be named, which is not a finding. When no engine can be named for any of the five, the column is left out. Your own site is listed alongside the rest, tagged Your site in place of a source type, and is frequently the biggest single source: that is worth seeing rather than hiding, because a position resting mostly on your own pages is more fragile than the count suggests. Source Balance on Citations is where that gets measured properly, as a share. The card covers the latest week. The footer counts how many sources gained or lost share and names the period it compared: the previous week, the previous 30 days or the previous 90 days, the longest your history allows, which is the same comparison Which sources to target makes on Citations. That count covers every source you have, not only the five listed, so it is normally the larger number. View all citations opens the full picture.

While a project's onboarding is still running, Overview and every other dashboard page show the onboarding card instead of partial numbers. That is intentional rather than a limitation: half-populated figures invite conclusions the data cannot yet support. If onboarding stopped before it finished, every page shows a card saying where it stopped instead. If your free plan has ended, that notice comes first. See Getting started.

Reading the GEO Score#

The headline number is the GEO Score, an overall measure of how discoverable your brand is to AI engines on a 0 to 100 scale. Three things feed it, and the card's own tooltip names them: AI Discovery Reach, meaning how often AI engines surface your brand; Readiness, meaning how well your site is set up for AI engines to crawl and understand, taken from your page audit; and Brand Sentiment, meaning how positively AI engines describe you.

Discovery reach is the biggest contributor, which has a practical consequence the tooltip also states: the GEO Score and the AI Discovery Reach card usually move together. When they part company, the mover is one of the smaller ingredients, and that is worth noticing because the response differs. A score dragged down by readiness is fixed on your own site through Smart Audit; one dragged down by sentiment is worked through Sentiment and its aspects; low reach itself is the longer game of topics, prompts and gaps.

The score renders green from 80 up, amber through the broad middle, and red below 20, so the colour is a coarse verdict at a glance. Treat the bands loosely: the number is comparable against your own history, not against other companies. The score is for the latest week's runs and the change beside it is in points against the previous week. If the previous week had no runs there is nothing to compare with, so no change is shown, which on a monthly schedule is most weeks.

AI Discovery Reach deserves its own sentence, because its definition is precise: the share of the latest week's prompt runs where an AI engine mentioned your brand, where a run is a single prompt on a single engine and every successful run counts toward the total whether you appeared or not. It is the same number as the latest week on the Visibility screen's reach trend, and its change is in points against the previous week. It tracks how often you surface, not where you rank, so you can place well in the answers you do appear in and still have low reach. Rank is Discovery Rank's job.

Interpreting results#

Read direction rather than value. A reach of 34% tells you very little by itself, because there is no universal number that counts as good and no benchmark that survives being applied across categories. The same card rising steadily across runs tells you something real, which is that whatever you have been doing is landing.

Brand Sentiment names the week in one word and draws the Sentiment screen's Current Sentiment Share beneath it as a bar of positive, neutral, negative and mixed answers. The word is Neutral when half or more of the answers carry no opinion. Otherwise it follows how far positive answers outnumber negative ones, so every week lands in exactly one band:

• Positive leads by 30 points or more: Positive.
• Positive leads by 10 to 29 points: Mostly positive.
• Within 10 points either way: Mixed.
• Negative leads by 10 to 29 points: Mostly negative.
• Negative leads by 30 points or more: Negative.

A mixed answer is both positive and negative, so it moves neither side. The change beside the word is the positive share, in points against the previous week.

Treat a single week's movement as noise until proven otherwise. AI answers vary between runs for reasons that have nothing to do with you, which Reading trends covers properly. If something moved sharply since yesterday, ordinary variation is the more likely explanation, and it is worth ruling out before anybody sends an alarmed message.

Let the outlier decide your next click. If one engine's row in the LLM Performance Index is out of line with the rest, or one card moved while three held still, that is where the next hour goes, and the rest can wait until it does not.

Best practices#

• A page you can skim in under a minute and correctly conclude that nothing needs you today.
• Movement concentrated where you have actually been working, which is about as close to attribution as a dashboard can get you at a glance.
• The GEO Score and reach moving together. When they split, you have a readiness or sentiment story worth reading before it grows.

Next steps#

Where you go from here depends on why you opened it:

1. The score moved and you want to know why. Check which ingredient moved with it: reach points at Visibility and Gap analysis, readiness at Smart Audit, sentiment at Sentiment.
2. You want work to do. Skip Overview altogether and go to Gap analysis, which is the screen that produces tasks rather than readings.
3. You are checking whether last month's work paid off. Go to Tracked actions and read the movement against what you actually did, rather than trying to reconstruct it from memory.
4. You are reporting to somebody else. Use Brand reports rather than screenshotting cards, since a report covers a defined period and states its own basis, which a screenshot cannot.

What to keep in mind#

Overview summarises, and that is all it does. Every number here appears in more useful form on the screen it came from, and no decision should be made from a card alone.

The GEO Score is Seerly's own composite. It is built for comparison against your own history, and there is no industry table of GEO Scores to rank yourself in.

It covers the engines your plan tracks and the prompts you have configured, in common with everything else in the product. See seerly.app/pricing.