Claims in this piece are observable and methodological — no client or benchmark data used.
A quality director reads through the ten moves that improve how AI assistants find and describe a contract manufacturer, agrees with every one, and then asks the only question that matters on a Monday morning: with one marketing person and a plant to run, where do we actually start? Ten changes is not a plan. It is a list. The full set lives in Ten Practical Ways to Improve How AI Finds, Understands, and Recommends Your CDMO; this piece is about the order to do them in.
The instinct, almost always, is to begin with the website: commission a new capability page, refresh the homepage, maybe a glossy PDF. That instinct is usually premature, and sometimes wrong. The better first move costs nothing and builds nothing.
Start by looking, not building.
Look before you touch anything
You cannot prioritise a gap you have not seen. Before a single page is rewritten, find out what the major assistants actually say about you today. Ask ChatGPT to recommend CDMOs for your exact core capability, say high-potency oral solids for a US-regulated market. Put the same question to Gemini and Perplexity, and to Claude and Copilot if you have access. For each, note two things: do you appear at all, and is what it says about you correct?
Treat this as observation, not a verdict. Answers vary by platform and shift over time, so no single run settles anything. But it is the map you cannot plan without, and the first read is usually sobering. The common result is not silence; it is being described as something you no longer do.
Fix what is wrong before you add what is missing
The six checks behind those ten moves are not equal in urgency, and treating them as a sequence is where most of the early return hides. A wrong fact costs you actively, because a contradiction reads as risk and risk disqualifies you before anything else is weighed. A missing fact is merely a gap. So correction comes before creation. (The six checks themselves are explained in the cornerstone; here they are simply put in order.)
Where to start: the order that returns the most trust for the least effort
- 1. Look first. Run the five-platform check and record what each assistant says about your core capability.
- 2. Correct what is wrong. Reconcile contradictions across your site, directories and public records, and fix inaccuracies. (Coherence and accuracy.)
- 3. Make your specialisation unmistakable. The single edit most likely to get you shortlisted for a specific prompt. (Specificity.)
- 4. Then fill the gaps. Add the fuller picture — scale, markets served, evidence, verifiable certifications — as the longer project. (Completeness and supportability.)
Correcting a three-year-old directory entry that still lists a discontinued line is faster, cheaper and higher-return than producing new content, and it is a trust move rather than a marketing one. A buyer extends trust to a supplier whose story holds together everywhere they check. Completeness comes last not because it does not matter, but because it is the largest effort for the most incremental gain, and it is wasted if an unresolved contradiction is still quietly disqualifying you upstream.
When did we last look at what an AI actually tells a buyer about us, or are we still polishing a website and hoping?
What if we can only do one thing this quarter?
Then correct the single most visible inaccuracy about your core capability. Not add, correct. The cheapest win in AI visibility is almost always removing a contradiction, because it takes away a reason to doubt you without requiring you to prove anything new. A CDMO that reconciles one conflicting capability claim across its site and its two most-cited directory listings has done more for how an assistant describes it than a quarter of new blog posts would.
Where to go next
Emerivo is a specialist advisory that helps pharmaceutical manufacturers and research organisations (TPMs, CMOs, CDMOs and CROs) understand and improve how AI assistants find, understand and recommend them. The ten moves are the full picture; this is where to begin. For the complete framework, read Ten Practical Ways to Improve How AI Finds, Understands, and Recommends Your CDMO, and when you want the gap mapped and prioritised for you across ChatGPT, Gemini, Perplexity, Claude and Copilot, that is what an AI Discovery Audit™ is for.
Frequently asked questions
Control is exactly why it feels like the place to start. But an assistant is not reading only your website; it is reconciling your site against directories and records you may have forgotten you have. If those disagree, a perfect homepage does not resolve the conflict, it just adds a third voice. Find the conflict first.
No one can tell you, and anyone who promises a date is guessing. Assistants refresh on their own schedules and differ from each other. What you can control is whether the information they eventually read is accurate and consistent; the timing is not in anyone’s hands.
No. The first step costs nothing but an afternoon of running the check yourself. An outside audit earns its place when you want the gap mapped systematically across all five platforms and turned into a prioritised plan rather than a to-do list, which is what an AI Discovery Audit™ is built to do.