Skip to content
SilktideHelp

How Presence works

Presence is built around a simple idea: are arenas, are contenders, and a mention is a contender observed in an arena.

Everything else - markets, prompts, scheduled measurement, traits, briefs, and fact checking - hangs off that model.

The hierarchy

Presence has two levels: your account, and the markets inside it. The account holds what every market shares. A holds everything one team curates and reads about one competitive space.

Account
├── Brands (yours and your competitors, as a tree of parents and sub-brands)
├── Regions (the places and languages questions can be asked from)
├── Platforms, how often you measure, and your plan
└── Markets (one competitive space each - compared, never added up)
    ├── Topics (an arena can be shared with other markets)
    │   └── Prompts (questions asked of AI assistants)
    ├── Brands here (which are yours, and which are competitors at which tier)
    ├── Regions here (which of the account's regions it is asked from)
    ├── Traits (the patterns claims are grouped into)
    ├── AI report, alerts, and tasks
    └── People (who can view it, and who can manage it)

Measurements (recurring runs of every active prompt, in every region, on every platform)
├── Responses (each answer, stored verbatim)
├── Mentions - who was named, in what order, in what tone
├── Sources - which URLs were used as evidence
├── Ads - paid placements some assistants serve beside answers
└── Claims - what was said about each brand

A few rules keep the picture simple:

  • A brand's part depends on the market. The same brand can be yours in one market and a competitor in another.
  • A topic in several markets is measured once. Each market reads the same answers through its own brands.
  • Markets are compared, never added up. There is no market above the others. A holding company that wants a picture of its parent brand gets an ordinary market with the parent brand as yours.

Each market's strategy brief, written by its AI optimizer, reads that market's measurements and proposes tasks. Those tasks belong to the market, and also appear in your account's task list beside the rest of your work.

Two things sit alongside this rather than inside it:

  • Accuracy is a parallel track on the brand itself: you declare facts, Silktide asks derived questions, and grades whether assistants agree. It does not feed share-of-voice rankings.
  • Traffic is first-party visit data from Analytics on your websites. It is the payoff of being mentioned: whether search engines and AI platforms then sent people.

From setup to standings

  1. Record brands - your own names and the competitors you care about, with aliases so matching stays reliable. They belong to the account, so every market can use them.
  2. Create markets - each market is one competitive space: which brands are yours there, who you are compared against, and which of the account's regions it is asked from.
  3. Define topics - arenas such as "undergraduate nursing programs" or "smartphones". Every topic is in at least one market, and markets that overlap can share one.
  4. Add prompts - the questions Silktide will ask. Most are generated for you, and every brand and market watching that topic shares them.
  5. Measure on a schedule - each active prompt is asked of each enabled assistant in each region; answers are stored verbatim and analyzed.
  6. Read standings - visibility, position, sentiment, citations, and traits for one market at a time, as trends rather than snapshots.
  7. Act - read the market's strategy brief, triage the tasks it proposes, and re-measure.

One pass, many numbers

Everything in Presence comes out of a single read of each stored answer. Silktide does not run separate scans for mentions, sentiment, citations, and traits - it reads the answer once and records all of it.

That is why the screens agree with each other, and why every headline number can be traced back to specific answers. Any figure you distrust can be drilled into until you are looking at the actual words an assistant wrote. Answers is where that trail ends.

Shared measurement

Prompts live on topics, not on individual brands or markets. Ten brands watching "building permits" cost the same to measure as one, and so do two markets that share the topic - Silktide asks the questions once and credits mentions to whoever appears in the answers.

This also keeps comparisons honest. Apple and Samsung are scored from the same smartphone questions, not from two private question lists that quietly drift apart.

Seen, tracked, and in a market

Three populations of names are easy to confuse, and the difference explains several screens.

What it means
SeenA name an assistant used in an answer. Silktide records everything it spots, which runs to thousands. Search them all under Suggested brands on the Brands screen.
TrackedA name you have added as a . Tracked brands are the ones with metrics: share of voice is divided among them, and claims are only extracted about them.
In a marketA tracked brand you have given a role in a specific market: one of yours, or a competitor. Those roles decide who appears as rows and columns on that market's screens.

So a rival you have never added is visible as a discovered name but has no metrics; a rival you track but give no role in a market still counts in the arithmetic but stays off that market's screens.

Account scope

Presence data is account-scoped, not website-scoped. Mentions are facts about a brand, not about a page. Citations can be traced back to a website when the cited URL belongs to a site in your account, but you do not configure Presence "inside" a single website the way you configure a website's modules.

Last updated

Was this page helpful?