Gå til indhold
SilktideHjælp

Getting started with Presence

Presence is not a report you read once. It is a loop: measure what assistants say, get a plan, do the work, measure again. This page walks the loop once, in order, so the rest of the help has somewhere to hang.

Budget about twenty minutes of your attention for setup, then a day or so of waiting while the first measurements come in.

1. Create your first workspace

The first time you open Presence you are asked to create a . You cannot skip it, and you should not want to - a workspace is the thing everything else is scoped to.

A workspace is one brand competing in one arena. Not your whole company; one competitive picture that one person could reasonably own. "UK business banking" is a workspace. "Undergraduate admissions" is a workspace. "Everything we do" is not.

The wizard starts the same way for everyone:

  1. Website. Your website address, or the brand's name if you do not have the URL to hand. A name is looked up; if Silktide cannot find the site, it asks you to type the address. Silktide checks that the page is real, then reads it.
  2. Location. While the site is read, a globe turns under a short wait. When the draft lands, it flies to the country we found. The card over it holds the brand name, a short description, language, and country, filled in from the page. Language follows the country unless you change it yourself. A local business can add a city and choose whether the first region is that city, its state, or the whole country. Edit anything that is wrong; most people just confirm. You can add more regions later under Settings → Regions if your plan allows.
  3. Workspace. Only appears when the analysis found several genuinely distinct workspaces. Pick the one to start with. You can add the others later.
  4. Competitors. Silktide's best guess at who you are up against, each with a checkbox. Untick anyone who does not belong and add anyone it missed - a name is enough, a website is optional. Every result in Presence is measured against this list, so it needs at least one name. Tracking a competitor is free.
  5. Topics. It proposes - the arenas inside the workspace - and writes for each, often already filling in while you are still on the earlier steps. The first topic opens so you can read the questions; the count on a row expands the rest. Untick topics you do not care about. Edit any prompt that does not sound like a real customer. When your allowance has room after the workspace's own topics, the first few competitors get a reputation topic of their own ("is X legit", "alternatives to X"); turn off any you do not want.

Topics is the last step. It shows your prompt allowance and how much of it this workspace will use. Prompts are the unit of measurement and therefore the unit of cost: every active prompt is asked of every assistant you enable, every time measurement runs. The plan spends what you have left, in order. If nothing fits, it says so instead of creating a workspace with no questions. Start measuring creates the workspace, topics, and prompts, and queues the first measurement run.

A reload resumes the draft. You can create more workspaces later from New workspace, and one brand can front several.

2. Wait for the first answers

Measurement is not instant. Silktide has to ask every active prompt of every enabled assistant, then read every answer and extract what it found. The first run takes a while; after that it repeats on a schedule.

Until that first pass finishes, Presence holds the workspace on a waiting screen instead of showing reports. Early numbers from a handful of answers would mislead. You can leave the page; measuring continues in the background, and Silktide emails you when the first scan is ready.

While you wait, it is worth understanding what "measured" means. Silktide asks each question through each assistant's own interface, stores the answer word for word, then reads it for:

  • every brand named, and in what order
  • the tone of each mention
  • every source URL used as evidence
  • every claim made about a brand, grouped into

That single pass is where every number in Presence comes from. Measuring visibility covers cadence, assistants, and why an individual answer is close to meaningless.

3. Cultivate the three lists

Presence is only as good as what you point it at. Three lists need your judgment, and they repay it.

Topics are the arenas. Good topics have a qualifier: "undergraduate nursing programs" beats "nursing". Each subject topic belongs to exactly one workspace, and each one costs measurement, so a short list of arenas you genuinely contest beats a long list of aspirations. Manage them under Settings → Topics.

Prompts are the questions. Most of them deliberately do not name you - "what is the best CRM for a small charity?" is worth more than "is Acme CRM any good?", because the first is where you get discovered. Edit the generated ones until they sound like your customers, add the questions you know people ask, and delete the ones that turn out to be too narrow or duplicated.

Competitors are who you are scored against. Add the rivals you care about under Settings → Brands, which lists your own brands and the competitors beside them. Your main brand's profile is under Settings → Brand profile. You do not have to guess: the Brands screen suggests names that keep appearing in your own answers, and you accept or ignore them.

Traits are the fourth list, and this one mostly cultivates itself. Silktide finds recurring traits in what assistants say - "opaque pricing", "strong accessibility support", "built for developers" - and groups individual claims under them so you can compare who owns which trait. Your job is pruning: merge duplicates and hide the ones that are noise. See Traits.

4. Read the standings

Once answers are flowing, the Overview screen is the front page of the workspace: the latest plan, the work in front of you, and where you stand on AI and search. The full numbered report lives on Visibility.

Two habits will save you from wrong conclusions:

  • Read trends, not days. Assistants are non-deterministic. A number that moved yesterday probably did not mean anything; a line that has been falling for a month certainly does.
  • Read visibility and share of voice together. Visibility is how often you appear. Share of voice is how much of the conversation you own. They move independently, and the gap between them is usually the finding.

5. Get a plan

Numbers do not tell you what to do. That is what strategy briefs are for.

A brief is written by an AI strategist that reads your measurements - your prompts, the answers assistants gave, who beat you, what was used as a source - and argues from that evidence rather than from general advice about your industry. It tells you who is winning the workspace and why, and it recommends specific work.

Run the workspace brief first. Once you know the shape of the problem, run topic briefs on the arenas that matter most, or commission a brief that answers a question you write yourself.

6. Triage the tasks, then do them

Every recommendation in a brief arrives as a task, proposed rather than imposed. On Overview, What to do lists those tasks so you can accept, decline, or prioritize them without leaving the front page. The Tasks area (and Presence Tasks) shows a Suggestions awaiting review strip when proposed work is waiting.

You decide:

  • Accept it and it joins your task list as real work, to be prioritized, assigned, estimated, and closed.
  • Decline it and it is cancelled - kept on record so a future brief knows you considered it and said no.
  • Edit it first if the recommendation is right in substance but wrong in detail.

If you have connected Linear, an accepted task can be published as an issue in your own tracker, so the work lives where your team already works.

7. Measure the difference

This is the step people skip, and it is the one that makes the rest worth doing.

Run the brief again after the work lands. It does not overwrite the last one - each run is a new edition, judged against the same fixed findings as every other edition, and the new edition knows which of its predecessor's recommendations you completed, declined, or left open. So it leads with what changed, tells you whether the work moved anything, and carries forward what is still outstanding instead of proposing it again from scratch.

That is the loop closed. Everything after this is repetition, and repetition is the point.

Where to go next

Sidst opdateret

Var denne side nyttig?