A meaningful share of your launch audience will never see the launch. They will ask an assistant - "best tools for X", "how do I solve Y" - and read an answer assembled from whatever sources the model trusts, which as of launch morning do not include the thing you just shipped. Launch playbooks cover press, social, email and paid; almost none cover the answer engines, even though the answers are where a growing slice of category discovery now happens. This walkthrough is the launch program with that surface included, run on the AI visibility and content agents.
The launch surface nobody briefs
The structural problem: assistants compose answers from already-trusted sources, so on launch day the category questions your product should own are answered from a corpus in which you do not exist. Nothing about the launch moment fixes that - the press release is one document, and assistants award AI citations based on retrievability, structure and corroboration rather than recency alone. The consequence is a timeline shift: the citation campaign starts six weeks before the announcement, which is the part launch checklists miss.
Six weeks out: the baseline audit
The program opens with the AI visibility audit scoped to the launch: the visibility agent enumerates the question set the launch should win - category queries, problem queries, comparison queries, the "best tools for" family - and records what each major assistant currently answers, who gets cited, and where you appear (usually: nowhere). The audit produces the target list: which questions are winnable, which citations are displaceable, and which competitor sources currently own the answers. It also produces the baseline every post-launch claim will be measured against, which is what makes "did the launch move the answers" a number instead of a feeling.
Four weeks out: content built to be cited
| Asset | Built for |
|---|---|
| Category definition page | The "what is X" queries - liftable definition up top |
| Comparison and alternatives pages | The queries where buyers weigh options - honest tables |
| Spec and capability answers | The precise questions assistants quote verbatim |
| Problem-solution guides | The upstream queries that precede category awareness |
The content agent drafts the set following the LLM SEO citation-format rules - answers stated early and liftably, structure assistants can parse, claims sourced - grounded in the audit's target questions rather than generic keywords. Publishing four weeks early is deliberate: retrieval systems need time to find, index and begin trusting new sources, and content that exists only on launch morning competes with nothing in its favor. Everything routes through the editorial QA gate like any other content - launch pressure is historically when quality gates get skipped, which is the argument for gates that do not depend on anyone's discipline.
Launch week: the technical pass and the push
Launch week adds two motions. The technical GEO pass verifies the machinery - crawlability for AI user agents, structured data, clean heading hierarchies, llms.txt - so nothing mechanical blocks retrieval of the pages just published. And the coordinated push - press, partner posts, community threads, the launch content itself - matters double here, because corroboration across independent sources is one of the signals assistants weigh: the same claims appearing on your pages, in coverage, and in discussion threads is what moves a fact from "one site says" to "sources agree".
After: tracking whether the answers moved
The steady state after the launch window closes is the standing generative engine optimization program: citation share tracked weekly beside share of voice, gaps feeding the content queue, and the next launch starting its audit six weeks out as routine rather than discovery. The AI visibility solution page maps the full program; the launch version above is that program pointed at one date.
Frequently asked questions
When should AI citation work start for a launch?
Six weeks out: baseline audit first, citation-built content published around four weeks out, technical pass and coordinated push at launch. Assistants need time to find and trust new sources, so launch-day-only efforts miss the window.
What content wins launch citations?
Assets built for how assistants select: liftable category definitions, honest comparison tables, precisely-stated spec answers and problem-solution guides - grounded in an audit of the actual target questions rather than keyword lists.
How fast do AI answers change after a launch?
Weeks, unevenly by assistant. The program tracks weekly against the pre-launch baseline and aims a second content wave at the gaps that did not move - expecting overnight change misreads how retrieval works.
How is success measured?
Against the baseline: share of target questions where you now appear, citations won and from whom, and movement per assistant - the audit’s before-state making the after-state a real number.
Sources
Every playbook on this blog ships as a runnable mission.
Open a workspace and the playbook library is waiting - describe the outcome and the agents carry it end to end, on your plan's monthly credits.