You start invisible, not penalised
A new site has no corroboration, which means models have nothing to weigh. That is a different problem from being ranked poorly, and the response is different: you are establishing existence and category before competing on preference.
Weeks 1–2: be unambiguous about what you are
One canonical description, used identically everywhere, plus organization schema that agrees with it. This is nearly free and it is what makes you resolvable as an entity rather than an unknown string.
Weeks 2–4: answer the questions you can answer best
Not the highest-volume terms — the specific ones where you can say something concrete. A new site cannot win broad category questions and can absolutely win narrow ones, which is how to choose what to write.
A new site wins narrow questions first. Competing for the broad ones on day one is a way to publish for nobody.
Weeks 4–8: state the facts competitors avoid
Pricing ranges, timelines, who you are wrong for. These require no authority to be quotable and they are the fastest route into answers where established competitors have refused to say anything.
Weeks 6–12: get described elsewhere
Profiles, directories, partner pages, being interviewed. Corroboration is the slow lever and starting it early matters more than doing it perfectly — it is what eventually decides recommendation.
Measure from the start
Set your fixed prompt set in week one so you have a baseline. Without it you cannot tell in month three whether anything worked, and there is no ranking to fall back on.