Skip to main content

​
What problem does it solve?

Buyers now ask an assistant before they open a search engine, and the assistant offers two or three options. If you are not one of them, you are not considered. Three things stop most teams acting on that:
  1. The numbers are opaque. Share of voice, citation rate, sentiment: it is hard to tell which one should change what you do this week.
  2. The tools need an operator. Dashboards say what is wrong and leave the fixing to you.
  3. Nobody has the time. Even a correct recommendation sits untouched, because acting on it is a week of work nobody scheduled.
Attensira takes the job. Agents do the work, show every piece of it, and measure what it did. You set the direction and approve what goes live. See How we work for who does what.

​
How the loop works

1

You set goals

At signup the agents read your site and propose goals. The first is usually to be seen at all: named in a number of the questions that matter to you, on the assistants you chose, by a date. Each goal has a metric, a baseline, a target, a date and a pace line it is read against.
2

The Chief of staff plans and staffs it

The Chief of staff proposes projects, each a crew of agents on one area of work, such as your blog or the comparison questions. You accept, change or remove them. You never define an agent or write a schedule.
3

The crews do the work

Pull requests to your site, articles, page rewrites, directory listings, community replies and outreach drafts. Every morning the Daily win plan also does up to three pieces of work toward the questions you want to win.
4

You approve what goes live

Anything that cannot be undone waits for your yes in the Inbox, along with the questions only you can answer and the few things only you can do. Work lands in your own systems: your repository, your CMS, Notion, your mailbox.
5

We measure, every day

Each day we put your prompts to the assistants you track and store the answers. Every change is read before and after it ships, and each goal is read against its pace line. A goal that falls behind is re-planned.

​
What we measure

A Prompt is one question, phrased the way a buyer would phrase it. Each prompt runs against each Model you track, and every Answer is stored and read for two separate things, for you and for each competitor you track: whether the brand was named, and whether the model attached a link to the brand’s domain. These are real answers from live model surfaces, not predictions and not a simulation.

​
Mention versus citation

A Mention means the answer named your brand in its text, with no link. A Citation means the model attached one URL pointing at your own domain. They are counted independently: an answer can mention you without citing you, cite you without naming you in prose, do both, or do neither. That distinction matters. A mention means the model knows who you are. A citation means the model sent the reader to your page. The two move for different reasons and are fixed by different work, so Attensira never collapses them into a single score. Share of voice is the share of successful runs that named your brand. Citation rate is the share of successful runs that cited your own domain. Competitor rates use the same denominator, so they will not add up to 100%.
A Source is a domain bucket across all citations in a window, including domains that are not yours. It tells you which sites the models lean on when answering your prompts.

​
What it does not do

  • It does not guarantee outcomes. We promise a volume of work for your budget and do not charge for work we don’t deliver. No one can promise what an assistant or a search engine will say. See Plans.
  • It does not claim revenue. We show what changed in the answers and in your traffic beside the work we shipped. We never claim a sale was caused by it.
  • No person on our side does client work. The agents do it, and a person is asked only for what a person has to decide or do.
  • It does not rank brands against each other. Your position counts where you came among the names you track in one answer, and there is no share-of-voice figure that sums to one.
  • It does not count an answer that names you without judging you as neutral sentiment. Sentiment reads only the answers that take a position on you.
  • Crawler traffic is a separate measurement. A crawler fetches pages and a model answers prompts; those are separate systems, read separately.