Brand map

Every other competitive screen assumes your list is right. This is the one where that assumption gets checked.

It is a map of the names AI assistants offer as answers to the same questions: your brands, your tracked competitors, and - the part no table can do - every company and product recommended beside them that nobody has added. A ranking computed over tracked brands looks exactly the same whether the list is complete or missing the market leader. Here the leader is a large dot with no logo, sitting next to yours.

Two pictures of the same thing

The switch beside the title chooses how the same set of names is drawn, because two quite different questions were being asked of one picture and it could only ever answer one of them well.

Explore - what is this name named alongside, and what are those names named alongside? One name in the middle and a branching picture you open a name at a time. This is where you go to follow something.
Map - what does the whole workspace look like at once? Every brand you are filtered to, in one still picture where position is a measurement. This is where you go to see the shape of it.
Table - the same names as a list, and the one to use if the pictures are not for you.

Both pictures size and color the dots the same way, and both let you click a name to read about it.

What you see	What it means
Size	How often a name was offered as an answer - not how often it was mentioned. A dot's width is proportional to that count, measured against the biggest name on screen, so a name offered twice as often is drawn twice as wide. The caption under each name states the number.
Color and logo	The brand's own color and favicon, for names you track. Where there is a logo it sits on a white circle ringed in the brand's color, because most logos are drawn for a white background and vanish on anything else.
A grey dot	A name nobody is tracking. There is no brand color to use, so the plain grey says "not yours" instead.

Sizes have a floor. A name offered twice would otherwise be a fraction of a pixel, and a dot nobody can point at is not on the map at all - so the very smallest dots are drawn slightly larger than truth to stay clickable.

Names are written wherever there is room for them, so a long one may be shortened to fit - and where things are tight, some dots go unlabelled altogether. Point at any dot to read its name in full, or click it to open a panel that names it and explains it. Nothing here is unidentifiable; the table lists every name as well.

Reading the explorer

One name sits in the middle, drawn with the handful of names it is most often offered beside, and each of those hangs off it on one line. That is the whole rule. A name's position says which name brought it and how strong that tie is, and nothing else - so opening something can only ever add to the picture.

What you see	What it means
A line	The name at the outer end is offered alongside the name at the inner end. Every name has exactly one of these, to whatever it hangs off.
Line weight	How strong that tie is. Heavier and darker is a name offered beside its parent far more often.
Distance along the line	The same measurement again: a name almost always offered beside its parent sits close to it, one that occasionally is sits further out.
A ring around a dot	It is showing its own list. Click it again to put that list away.
A very faint line across the picture	Something else lists a name that is already drawn elsewhere.

Every name is drawn once. Where a name you open is also listed by something that already has it on screen, you get a faint line to the dot that exists instead of a second dot. Two brands in one market share most of their neighbours, and drawing a shared name once per brand meant the biggest names were duplicated the most: on a real account one competitor appeared five times at full size, and a third of the dots on screen were copies.

Nothing already on screen moves when you open a name. Room for what a name might list is left beside it in advance, so its list arrives where you are already looking. Deep down a long branch that room runs out, and a list will lean toward whichever side has space - but it stays on its own line, off its own parent.

Choosing what is in the middle

The middle starts as your own brand - or the brand you have filtered to, if it is one. From there:

Click any dot to open it where it is. The middle does not change; the picture grows.
"Start here", in the panel, clears everything away and rebuilds the picture around that name instead. This is how you get into a workspace several hops from anything you track without dragging the trail behind you.
"Start from your brand", beside the title, goes back.

A tree has one root, which is why this picture has one name in the middle where the map has all of them. With eleven brands in the middle, everything they share has to hang off each of them, and the result was unreadable.

Reading the map

The map has no branches and does not open. It draws every name at once and puts the measurement into position:

What you see	What it means
Distance	How much of the same two names answer. Near means they keep being offered as alternatives to each other; far apart means they rarely compete.
A solid line	The two names answered the same questions. They compete.
A dashed line	The two names share questions without answering them - see Competing and being mentioned below.

Because distance is shared answers rather than similarity of any other kind, clusters form on their own: a group of names huddled together is a set of companies offered as substitutes for each other. That is usually a workspace segment, and occasionally it is one company appearing under several names (see below).

Where the explorer starts from one name, the map draws every brand the Brands filter names, each with a few of its own nearest - so this is the picture to read when the question is about the field rather than about one company in it.

Competing and being mentioned

Sharing an answer is not the same as competing for it, and the map draws the two differently.

When an assistant answers "which accessibility platform should we buy", it names the products it is recommending - and it also names the ticketing system they integrate with, the screen readers used to test, and the CMS the site runs on. Counted the same way, an integration listed in every vendor's feature table looks like the most tightly connected company in the workspace. On one real account, the strongest single relationship in the entire dataset was a web governance platform and Jira.

So each name in an answer is judged on whether it was offered as one of the answers or merely named while answering, and the map is built from the first of those:

Solid line - competing. Both names were put forward as candidate answers to the same questions.
Dashed line - mentioned beside. They share questions without being alternatives to each other.

Only the competing relationships count by default, and only competing names are drawn at all. "Include names you're only mentioned beside", beside the title, adds the rest: the ticketing tools, the CMSes, the assistants themselves. On the map they arrive on dashed lines and around the outside, because being named in the same breath as a workspace is not the same as being in it. In the explorer they are added to the end of each name's list, after its rivals, so turning them on cannot rearrange the competitive picture - it can only extend it.

Nothing is on a list somewhere. A name is context because of how it gets mentioned in your answers, so the same name can be a competitor on one account and scenery on another - which is the only way this can be right for a customer who sells against ChatGPT.

Why there are so few lines

Every pair of names that answers a prompt together is a connection, so twenty names carry up to a hundred and ninety of them. Drawn all at once they are a grey wash that says "everything is connected to everything", which is not a finding.

So neither picture draws them all. The explorer draws one line per name, to whatever it hangs off. The map draws each name only the ties close to its own strongest one, and never more than four, which leaves the shape of the workspace rather than a record of every pair. In both, lines get heavier and darker with strength, so a rivalry looks like one and an acquaintance does not.

Point at any name to draw its strongest ties, wherever the other ends happen to be, and their names appear too. That is the moment somebody is actually asking - and in the explorer it is how you see that two names on opposite branches are in fact close, which the branches themselves cannot tell you.

What the filters do

The Brands filter in the toolbar decides what the map puts in the middle. Its default of "All" draws your whole tracked workspace rather than one brand chosen for you, and narrowing it to two or three is how you get a head-to-head with the rest of the field out of the way. Narrowing it to exactly one also moves the explorer's subject to that brand, so it is a way of pointing at a competitor as well.

Everything else on screen arrived because it shares answers with what is in the middle. Each name brings the same number of its own nearest, decided by the workspace and by nothing else - not by how many other names are on screen, and not by the size of your window. The and date filters apply as they do everywhere in Presence, so both pictures are always drawn from the answers those filters describe.

Moving around

Either picture is a world with a camera on it rather than a diagram squeezed into the window. Scroll or pinch to zoom, drag to pan, and use the fit control in the corner to put the whole thing back on screen. Names appear beside their dots as there is room for them, so zooming in labels more of the workspace and shows how often each name was offered.

This is why the same workspace is the same shape on a laptop as on a large monitor. How much is drawn is decided by the workspace and by what you have opened - never by how big your screen is.

Distances are measured between the edges of two dots rather than between their centres, so a big name claims the space it occupies and the gap either side of it is real. That is what leaves somewhere for the names you open next to go.

Clicking a name

In the explorer, clicking a name opens its own list and tells you who it is. Its names spring out of the dot you clicked and settle around it, and nothing already on screen moves. Clicking it again puts them away without closing the panel. That is how you follow a thread: click the untracked name sitting beside you, and see what it gets compared against, and then what those are.

The camera comes with you. Opening a name settles the window on that name and what just arrived, and leans in a step so there is room to read it. It leans a little further each time, and stops well short of a close-up. Fit puts the whole thing back whenever you want it.

On the map, clicking a name only tells you who it is, because the map is one picture of one workspace and there is nothing to open. To walk outwards, switch to the explorer.

The panel is the answer to "who is that?", which a dot cannot give you:

Why it is here - the name whose neighborhood brought it in, unless you asked for it directly.
Where it turns up - which assistants named it, which of your topics, and when it was first and last seen in this window.
Most named with - its whole neighborhood, not just the part that was drawn. Click any row to read about that name instead; a row marked + is not drawn yet, and clicking it puts it there.
Recent answers - go and read one.
Also written as - every spelling that resolves to this name, which is how you tell one company written two ways from two companies with similar names.

The actions are in the panel too, for untracked names: add it as a competitor, merge it into a brand you already track, or ignore it. Start here is there as well, in the explorer, for anything that is not already the name in the middle.

Clicking is a step in the browser's history, so back retraces your path, and where you started, what you have opened and what you are reading about are all in the address bar - a picture you have built up can be sent to someone.

How much is drawn

The control beside the view tabs - "8 each", or whatever it currently says - decides how many names every name in the picture is drawn with. The explorer starts at eight and the map at five, because the explorer has one name in the middle where the map has your whole tracked workspace, so the same number gives two very different pictures.

Turn it up to see more of the field at once; turn it down for a picture you can read in a glance. It applies to what is in the middle and to everything you expand, so the picture stays the same density however far out you walk. There is a limit on the total, so a long trail of expansions cannot run away with the screen.

The table

The Table view lists the same names, strongest tie first. It is the accessible reading of the picture, and it carries the same actions the panel does - merging a name into a brand, adding it as a competitor, ignoring it.

Two count columns, because they answer different questions. Offered as an answer is what both pictures are drawn from. Answers is every mention including the ones in passing, which is the number the rest of Presence reports. A name where the two are far apart is being talked about far more than it is being recommended.

What this screen is good at

Finding rivals you are not tracking. A large unlabelled dot beside yours is a competitor your rankings currently cannot see. Add it and every other Presence screen starts including it.
Spotting one company appearing twice. Products and sub-brands are extracted as their own names, so you may see "Acquia" beside a tracked "Acquia Web Governance", or "axe-core" beside "Deque". Two names that are really one company is a merge waiting to happen, and the panel and the table are both where you do it. On the map they show up as two dots almost on top of each other; in the explorer, as two names side by side in the same list.
Finding what to write about. Turn on names you are only mentioned beside, and you get the tools, assistants and assistants your workspace is discussed alongside. A company that keeps coming up beside yours without competing with you is an integration page, a comparison, or a partnership you have not written yet.

What it cannot tell you

Being offered as alternatives to the same question is strong evidence of competition, but it is evidence rather than a declaration. An assistant answering "alternatives to X" will list companies that only partly overlap, and two names can be offered together because they are usually bought together rather than instead of each other. The map states how the assistants treat them; whether that matches how the workspace actually behaves, you know and it does not.

Nor is nearness a score, in either picture. It says two names answer the same questions, not that either wins them - read Heatmap for who is named more often and Visibility for who is named first.

Brands and competitors - adding a name here as a tracked brand or an alias
Heatmap - every brand against every arena
Metrics at a glance - position and the rest of the measured set
Prompts - the questions these answers come from