A study ends with a list: every source the models cite in your category, ordered by who can actually move it. Most teams read that list, agree with it, and then nothing happens, because the list is not a plan and nobody in the room owns it.
This is the service that works through the list.
Why this is a separate service
Measurement and remediation are different jobs, and pretending otherwise is how agencies end up selling optimization they never verified. So we keep them apart on purpose.
The study is the instrument. It tells you where you stand, against whom, and which sources decide it. Our measurement platform, answer.show, is explicit about where it stops: the project does not fix your site, does not commission editorial content, and does not negotiate with retailers. The report says which of those things would move the number, and who owns it.
That sentence is the specification for this page. The study hands you the list. This service works the list.
The three buckets, and where the work lives
Every source list sorts into three buckets, and the work is different in each.
You control
Sources you control
The smallest bucket, and usually the one teams over-invest in.
Your own site, your product pages, your documentation, your help centre. Real work here is unglamorous: make the claims specific, make them findable, make them consistent with what the retailer page says. Google's own guidance is that optimizing for generative AI search is still SEO, and we do not pretend otherwise.
You can influence
Sources you can influence
The biggest bucket, the one the models cite most, and the one nobody owns internally.
Retailer and marketplace pages, distributor catalogues, directories, review platforms, trade media, the communities where your category actually gets discussed. It is where most of the work in this service happens: rewriting the retailer page that three assistants quote, correcting the spec table that propagates a wrong claim, earning the mention in the trade outlet that keeps showing up in the citations.
You do not control
Sources you do not control at all
You cannot edit these.
Regulators, standards bodies, consumer organisations, independent press. But you can stop being absent from the conversations that feed them, and you can know what they say about you before a buyer does.
The report tells you which bucket each source falls in, how often it appears, and who owns it. The work starts from that order, not from a generic checklist.
What the work actually is
Scope is set per engagement against your own source list, so no two look identical. These are the recurring pieces:
Influence
Retailer and marketplace pages
The pages an assistant quotes when someone asks where to buy. Claims, specs, categories, titles, the answer to "is this the one for sensitive skin". Often the single highest-leverage item on a consumer brand's list, and almost always out of date.
Influence
Editorial and commissioned content
Not blog volume. Specific pieces that answer the specific questions your prompt bank showed the models struggling with, placed where the models already look.
Influence
Earned media and trade presence
The outlets that keep appearing in your citations, approached because they appear, not because they are on a generic media list.
Influence
Review and community presence
Where your category gets discussed and cited, handled as the public record it is — without astroturfing, which corrupts the same measurement we are trying to move.
Control
Your own properties
The site work the report actually justifies, and no more.
Handoff
Briefings for the teams who already own these
Sales, trade marketing, ecommerce and PR each hold a piece of this list already. Part of the job is handing each one their slice in a form they can act on.
The work is not to fix the page. It is to be accurate and present in the conversation around it.
What we commit to, and what nobody can
We commit to two things.
We commit to
- A prompt bank that is complete and useful for your brand, whatever your industry
- A run that produces useful findings
What nobody can guarantee — us included
- That the actions you take afterwards will move the numbers in a later measurement
The method is public precisely so you can judge the findings, not the promise.
That applies with full force here. Third parties write most of what an assistant says about your category, models change, and competitors are working too. What this service commits to is that the work is done on the sources the evidence points at, that every action is traceable to a cited source in your own report, and that the next run reports honestly whether anything moved — including when the honest answer is that nothing did.
If nothing moved, the report says nothing moved. That is the same standard, applied to our own work.
How it runs with monitoring
- The study sets the baseline
- This service does the work
- Monitoring re-runs the same frozen prompt bank
Monitoring re-runs the bank afterwards, so the second measurement is comparable to the first.
The order matters and it is not negotiable: no remediation before a baseline exists. Without the baseline there is nothing to compare against, and any claim about improvement is decoration. A comparison of two runs still does not prove causation — it shows movement, in a category where other people are also acting.
When not to buy this
You have no baseline yet
Start with the study. This service reads from its output and cannot run without it.
Your source list is dominated by sources nobody can move
If the answers in your category are decided almost entirely by a regulator, we will tell you, and the honest recommendation is a smaller scope or none.
You want a guaranteed position
Nobody controls what an assistant recommends. A provider promising positions in ChatGPT is selling something they do not control.
The decision the work would inform does not exist yet
Bring the decision first. It sizes the work.
Delivery
The priority and the standard are ours. The report’s ordered source list sets the order of the work, and every action traces back to a cited source in your report, so the next run measures what it named.
We execute directly
The work on the pages you control or can influence: retailer and marketplace listings, commissioned editorial, and the fixes on your own site the evidence justifies.
We coordinate, rather than do
The pieces that depend on a relationship only you hold, or on a specialised firm: earned media and press, the commercial relationship with your retailers, and regulated categories. There we write the brief, set the standard and review the result; the execution runs through your team or through editorial and PR partners.
Your own teams keep
What is already theirs: sales, trade marketing, ecommerce and PR receive their slice of the list in a form they can act on.
One line we do not cross
No astroturfing and no paid reviews, because they corrupt the same measurement this work is trying to move.
Questions
Is this just SEO with a new name?
Part of it is SEO, and we say so. Google's documentation is explicit that optimizing for generative AI search is still SEO, and it names llms.txt, chunking and special schema as myths. What is different is the targeting: the work is aimed at the sources your own measurement showed the models citing, in the order the evidence set, rather than at a domain-wide checklist.
Can I buy this without the study?
No. Without the frozen prompt bank and a first run there is no source list, no priority order, and no way to tell later whether anything moved. It would be optimization sold before measurement, which is the thing this agency exists not to do.
Do you do the work, or tell my team what to do?
Both are on the table, and the split is agreed per engagement. Some teams already own the retailer relationship and need the priorities and the briefings. Others need the work done. The report is the same either way.
What if the report says the biggest lever is something you cannot do?
We say so, and we scope around it. The source list frequently points at a regulator page, a retailer relationship you own, or an editorial outlet that only your brand can credibly approach. Naming that honestly is worth more than inventing work for its own sake.
Start with the decision, not the service
Before buying any remediation, name the decision it serves and the number you expect to move. Bring both to the conversation. If the study has already run, bring the report — the priorities are in it, and the conversation starts there.