The first days

Nothing needs starting by hand. Once the plan is active and the domain is verified, Vercatus builds the site’s topics, starts asking the answer engines about them and audits the site; until the domain is verified none of it starts (see Verify a domain). This page says what arrives, roughly in order, and what is worth doing in the first week.

Within the hour of verification: topics and the first answers

• Topics. Vercatus writes the site’s topics from the buyer’s side: what people ask an assistant when they choose, compare, fix or learn about something in your category. Each topic comes with phrasings written the way people type into an assistant, and those become the questions the poll asks. Your plan sets how many topics a site measures. See Topics and AI demand.
• Demand tiers. Each topic is marked high, medium or low demand from the search volume of its words. A topic with no figure keeps the tier the builder gave it until one is found.
• The poll. The first batch of questions goes to every available answer engine, and more follow through the day within the plan’s monthly allowance. See The poll.

The first days: Answer Share fills in

The Overview shows Answer Share from the first answers, with a wide range. While the poll is filling, it says how many answers it has so far and that ranges narrow as answers arrive. A share is never shown as zero because nothing was measured yet: an unmeasured share is a dash.

Competitors are discovered from the answers: every brand an answer names or cites is recorded, and the brands named most become the leaderboard. Pin the competitors you know, and hide brands that are not competitors.

Every day: the lenses and the plays

Once a day, for each site, Vercatus:

• reads the third-party pages the answers cited and records the brands each one names (Sources);
• looks up the searches the engines ran in the indexes they read (Search);
• asks a small share of the questions with search off (Memory);
• regenerates the plays from what the lenses found, and runs the acceptance checks that are due.

The Engine Briefs are asked once a week. AI demand estimates are computed daily and change when their inputs do. See Why and Plays.

When the domain is verified: the first audit

The site audit starts the moment the site is verified. It samples the site’s pages across its templates, reads each one as a browser and as each AI crawler, simulates whether your passages answer your questions, and writes findings. Page fixes and Foundations plays come from it. When it ends, the run report arrives by email; Reading a site audit explains it, and How a run works explains the stages.

The first week: what to do

1. Check the topics. On Topics, rename a topic, pause one you do not want measured, add the ones the builder missed, and add your own questions to a topic. A rebuild after the first audit uses what the audit learned about the business. See Topics.
2. Check the competitors. Pin and hide on the leaderboard or on Competitors, or add a competitor by name and domain. See Competitors, Sources and Briefs.
3. Connect your data. Search Console adds your real buyer searches to your topics, as questions private to your workspace and deleted if you disconnect, and its AI report to your results; Google Analytics and crawler traffic give Results its outcomes, and crawler logs date your page changes the day an engine’s crawler fetched them.
4. Decide on an experiment. Before you ship plays, you can hold some topics out on Results. Verdicts are stronger when topics are left alone on purpose. See Results.
5. Plan your first play. Open the top play, read its evidence, and set an owner. Home’s getting-started list marks this step done once someone has planned, started or dismissed a play.

Starting a run yourself

After the first audit, the schedule runs the site’s full and change runs for you. To start one by hand, press New run in the header (or ⌘K, New run). The dialog says what the run counts against before anything starts: which allowance, how much is left and when it resets, how many requests it makes and roughly how long the crawl takes. A refused run says why in one sentence and starts nothing. The same run can be started from the API or the CLI.