Day 121: Choose the Surface Before You Choose the Prompt
Before anyone buys five screenshots of the same prompt, decide what the buyer was trying to learn.
A CMO can ask a sensible question and still commission a weak piece of GEO research. The weakness is not the wording. It is the assumption that the same sentence, repeated across ChatGPT, Claude, Perplexity, Gemini, Google AI features, Google Search results, directories, review sites, and comparison pages, produces one coherent answer to one commercial question.
It does not.
A direct answer capture is one evidence class. A search result is another. A cited page is another. A directory profile, review surface, public comparison article, technical source inspection, or Google AI context is another again. They may all matter. They should not be treated as interchangeable channels in a coverage grid.
For CMOs, Marketing Directors, and founders, the procurement failure is simple: a supplier can sell platform breadth before anyone has defined the observation that would help the business decide. The better brief starts upstream.
Who is the buyer? What stage are they in? Which surface could plausibly shape or reveal that moment? What observation would count? What decision could the business make from it?
The surface is part of the question.
The same sentence can be several different studies
Imagine a founder-led B2B company buying a GEO baseline. Leadership wants to understand whether high-fit buyers can distinguish its advisory diagnostic from monitoring software, content production, technical SEO, and internal analytics work.
The team writes one reasonable sentence:
“Which partner should a B2B Marketing Director consider for diagnosing whether AI answers are sending prospects towards the wrong type of solution?”
That wording could support several jobs. It could test whether a direct answer-led surface describes the offer as advisory rather than software. It could inspect whether Google Search results expose the page that explains the offer boundary. It could examine whether a cited article, comparison page, directory, or review surface is introducing the wrong category. It could check whether public technical material gives crawlers and agents enough accessible context. It could help sales prepare for a buyer who arrives with a pre-formed shortlist.
Those are related jobs, but they are not the same job.
If the brief says only “run this across five engines”, the supplier may return a tidy deck and still leave the buyer unable to act. A screenshot from a direct answer does not tell the same story as a ranking page. A directory absence does not mean the answer engine refused the company. A comparison article mentioning a competitor does not equal buyer preference. A Google Search context should not be mixed with a non-Google answer capture as if both were one generic AI result.
The prompt did not fail. The research design failed.
Start with the buying situation
A useful surface-first brief begins with a scenario, not a platform list.
The buyer might be a Marketing Director at shortlist stage trying to compare advisory diagnosis with monitoring software. The buyer might be a founder checking whether the category is mature enough to fund. The buyer might be a procurement lead looking for risk and methodology language before approving a supplier. The buyer might be a sales-qualified prospect trying to understand whether the offer fits their team, market, or current stack.
Each situation asks for a different observation.
A shortlist question may need a direct answer capture under recorded conditions, plus the visible sources the answer uses where available. A procurement-risk question may need public methodology pages, Search results, cited sources, and reviewable claims. A fit question may need the offer page, integration notes, market boundaries, and sales-safe next steps. A category-understanding question may need search results, comparison pages, and answer-led explanations that show whether the problem is being framed in buyer language.
The team should name the commercial job before deciding where to look.
A compact brief can use six fields:
| Field | What to state before collection | Why it matters |
|---|---|---|
| Buyer role | CMO, Marketing Director, founder, procurement lead, sales leader, product marketer, or another named role. | The same observation has different value depending on who would use it. |
| Buying stage | Problem learning, category formation, shortlisting, fit check, procurement, risk review, or sales qualification. | The stage determines whether the answer should educate, compare, reassure, qualify, or route. |
| Surface or evidence class | Direct answer capture, Google AI context, Google Search result, cited page, directory, review site, comparison page, technical inspection, sales note, or another labelled source. | Keeps the observation from being collapsed into one generic AI channel. |
| Access context | Date, market, language, account state, location signals, visible sources, and any known limitation. | Prevents one capture from becoming a platform-wide claim. |
| Expected observation | Category named, offer boundary described, wrong substitute introduced, public source visible, review claim present, source missing, or next step unclear. | Defines what the team is actually looking for. |
| Decision informed | Clarify a page, challenge a buyer criterion, inspect a source, improve sales enablement, route to leadership, run a narrower baseline, or leave alone. | Stops research from becoming screenshot theatre. |
This is not bureaucracy. It is the minimum structure required to know whether the result belongs in a public-content brief, a sales note, a technical inspection, a procurement conversation, or nowhere yet.
Choose the surface that can answer the job
Surface-first scoping does not mean choosing one favourite platform. It means refusing false comparability.
If the job is to see how a buyer’s direct answer-led research frames a provider category, a retained answer capture may be relevant. Record the exact question, date, market, language, account or access state where known, visible sources, and limits. Do not turn it into attribution, ranking, or proof of buyer behaviour.
If the job is to understand why a direct answer may be plausible, inspect the public sources that could support it: service pages, methodology notes, comparison articles, directories, review sites, partner pages, technical documentation, or search-visible snippets. That is source inspection, not the same evidence class as the answer itself.
If the job is Google-specific, keep it Google-specific. Google’s own Search guidance says generative AI features in Google Search are rooted in core Search ranking and quality systems, with ordinary Search eligibility, usefulness, technical clarity, and people-first content still doing the work. Do not smuggle in a claim that llms.txt, special AI markup, arbitrary chunking, or over-focused structured data is a required switch for Google AI visibility.
If the job is procurement reassurance, the useful surface may not be an answer engine at all. It may be the methodology page a procurement lead can review, the public limitation language a founder can defend, the comparison asset sales can send, or the risk statement that prevents the buyer expecting deterministic visibility control.
If the job is category understanding, broad search and comparison surfaces may matter because they reveal the language and alternative routes buyers encounter before they know what to ask an answer-led system. Those surfaces are useful precisely because they are not direct answer captures.
The right question is not “Which engines did we cover?”
It is “Which surface can produce an observation relevant to this buyer job?”
A teardown of the weak brief
A weak brief says:
“Run the same prompt across ChatGPT, Claude, Perplexity, Gemini, and Google AI. Score whether we appear.”
It sounds efficient. It is easy to sell. It gives leadership a table.
But the table hides too much. It does not say whether the buyer was researching category education, supplier fit, procurement risk, implementation feasibility, or sales qualification. It does not preserve whether Google was being treated as Search, AI Overview, AI Mode, or a normal results page. It does not separate a direct answer from a cited source, a search result, a directory listing, or a review page. It does not say what the observation would allow the business to do.
A stronger brief says:
“For a UK B2B Marketing Director at shortlist stage, inspect whether answer-led and search-led surfaces preserve the distinction between advisory GEO diagnosis and monitoring software. Capture direct answer observations separately from Search results, cited pages, directories, and comparison pages. Record date, market, language, access context, visible source limits, and the public cues that may explain any misclassification. Use the result only to decide whether the offer boundary, sales enablement, or source inspection needs a narrower follow-up.”
That is less theatrical. It is also more useful.
The stronger brief does not promise that the company will be recommended. It does not claim a universal visibility score. It does not imply that all surfaces should agree. It does not average unlike evidence. It names a buyer, a stage, a surface set, an observation, and a decision.
That is what makes the work purchasable.
Do not buy breadth before coherence
Broad coverage can be valuable after the design is coherent. A mature programme may need repeated checks across several surfaces, markets, languages, buyer stages, and source types. A launch may justify a wider baseline. A regulated purchase may require procurement, Search, answer-led, review, and directory views to be kept in parallel.
The sequence matters.
Breadth after a clear buyer scenario can show how routes differ. Breadth before the scenario creates noise with a premium price tag.
The vendor theatre version sells the comfort of completeness: more engines, more prompts, more screenshots, more rows. The research-design version earns scope by explaining why each surface belongs. One surface may test a direct answer. Another may inspect Search visibility. Another may reveal a source gap. Another may expose a third-party category error. Another may be excluded because it cannot answer the job.
Exclusion is not a weakness. It is a sign that the brief has a standard.
A CMO should be able to ask a supplier:
- What buyer situation is this prompt meant to represent?
- Which surface can actually produce evidence for that situation?
- Are we looking at a direct answer, a search result, a cited source, a review surface, a directory, or a proxy?
- What access context and visible-source limits will be recorded?
- What claim will this observation not support?
- Which business decision could change if the finding appears repeatedly?
If those questions cannot be answered before collection, the scope is not ready. It may still produce artefacts. It will not produce a defensible baseline.
The procurement sentence
Put one sentence at the top of the GEO research brief:
We are choosing this surface because it can observe this buyer job under these conditions, and the result may inform this specific decision.
If the sentence cannot be written, do not buy the prompt run yet.
Rewrite the buyer scenario. Narrow the stage. Separate direct captures from proxies. Decide whether the work is about answer-led framing, Google Search visibility, source inspection, directory presence, review-surface claims, procurement reassurance, or sales qualification. Then choose the surface.
For CMOs, Marketing Directors, and founders, this is how GEO procurement becomes less performative. The goal is not to pretend the market is clean. It is to stop paying for messy observations as if they were comparable.
The same wording can be useful in several places. It just cannot mean the same thing everywhere.
Choose the surface before the prompt.
Then collect only the evidence that can survive the decision it is meant to inform.