Skip to content

Day 140: Sometimes the Best AI Answer Is ‘It Depends’

A buyer asks, “Which supplier is best for our AI visibility programme?”

The question sounds complete. It is not.

Best for a company that needs a one-off diagnosis or a managed monitoring service? Best for a team with authority to change its website or one that needs board evidence first? Best for a single market or a multilingual launch? Best when the problem is category confusion, product misinformation, weak comparison language, or inaccessible source material?

A confident list of providers may be fluent while answering a decision the buyer has not actually defined.

For CMOs, Marketing Directors, and founders, the aim should not be to make every buyer question produce an immediate brand answer. Some questions should produce a branch. Some should expose a missing input. Some should end with a better follow-up question.

“It depends” is not a weak answer when the decision genuinely depends on facts the buyer has not supplied. It is the start of a more useful one.

Confidence can hide a missing decision

An under-specified question invites the answer to fill in the blanks.

A request for the “best GEO agency” may omit the buyer's market, objective, available evidence, implementation capacity, procurement constraints, and need for ongoing operation.

Those omissions are not minor personal preferences. They can change the category of solution.

The buyer may need software, advisory work, technical implementation, product clarification, sales research, or no external purchase yet. Until the missing conditions are known, a definitive recommendation can collapse several legitimate routes into one apparently objective winner.

Every supplier named might be real and every capability quoted correctly. The answer can still be overconfident because the decision rule is absent.

A better public footprint does not merely provide more favourable claims. It explains which variables change the recommendation.

A generalised teardown: two answers to the same question

Consider a fictional B2B software company preparing to enter a second market. This is a generalised illustration, not a client result or an observed response from a named answer engine.

Its Marketing Director asks:

“Should we hire a GEO specialist or use our current SEO agency for the launch?”

A smooth answer might compare provider types, list benefits, and recommend a specialist because the project involves AI visibility. That sounds decisive. It also assumes the label “GEO” determines the work.

But the decision changes when four missing facts appear:

  • The current agency can handle crawl access, localisation, core Search work, and approved page changes.
  • The company does not yet know whether answer-led surfaces misclassify the offer in the target market.
  • Sales has no target-market objection record because the launch has not started.
  • Leadership wants evidence of a distinct commercial problem before funding another supplier.

Under those conditions, “hire a specialist” is premature. The responsible next step may be a bounded baseline or an internal check using existing capability. A specialist may become appropriate if the work reveals category confusion, answer-surface differences, competitor framing, or a decision problem that the incumbent cannot diagnose.

Change the facts again. Suppose the company already has repeated, bounded observations showing that important buyer questions place the offer in the wrong category, the current agency treats the problem as ordinary keyword coverage, and leadership needs an independent decision before committing launch budget.

Now specialist diagnosis may be the better route.

The supplier list did not change. The decision conditions did.

The useful answer is therefore not a universal winner. It is a branch:

  • if the problem is unconfirmed and existing capability can run a bounded first check, start there;
  • if a commercially important answer pattern is established and the team lacks the method to diagnose it, seek specialist help;
  • if the offer itself is unresolved, settle the business truth before buying visibility work;
  • if implementation capacity is absent, include delivery rather than purchasing another report.

That answer does more work than a ranking because it helps the buyer locate their situation.

Publish branch points, not a wall of caveats

The wrong response is to bury every page under defensive qualifications. Repeating “it depends” without explaining why merely replaces false confidence with fog. Useful conditionality names the few variables that materially change the route.

For a B2B offer, those variables might include:

  • Problem state: Is the issue observed, suspected, or still undefined?
  • Decision required: Does leadership need diagnosis, implementation, monitoring, assurance, or category education?
  • Existing capability: Can the current team or supplier perform the next responsible step?
  • Operating burden: Who can provide access, approve claims, implement changes, and maintain the result?
  • Market boundary: Do geography, language, regulation, plan, or buyer type alter fit?
  • Evidence threshold: What would justify moving from a small investigation to a larger commitment?

Instead of writing, “Our platform is the best solution for enterprise AI visibility,” write the conditions under which it is useful and the conditions that point elsewhere.

Instead of writing, “Book a diagnostic to find out why you are absent from AI answers,” state when a diagnostic is warranted, what a buyer can check first, and which unknown it is designed to resolve.

Instead of writing, “We support global teams,” name the markets, delivery conditions, or partner dependencies that change the answer.

This gives people and systems usable decision material. It does not guarantee that any answer-led surface will preserve the branch, ask the follow-up, include the company, or recommend the intended route.

Test for calibrated uncertainty

A useful GEO review should inspect more than whether the brand appears.

Choose a commercially important question that is easy to ask and impossible to answer responsibly without context. Record the question, surface, market, date, access conditions, and visible sources. Then inspect the shape of the response.

Does it:

  1. name the decision variables that matter;
  2. distinguish facts supplied by the buyer from assumptions added by the answer;
  3. narrow the plausible routes without pretending one is universally best;
  4. state what remains unknown;
  5. ask or imply the next discriminating question?

One captured response cannot show prevalence, buyer trust, causality, conversion, or revenue.

The test has a narrower purpose: does the available public material support a conditional answer to an under-specified buying question, or does it encourage false certainty?

A weak response names a winner before it knows the buyer's situation.

A stronger response might say:

“The right route depends on whether you need to establish the problem, implement a known change, or monitor an existing programme. Start by identifying the buyer question, market, available evidence, and internal capacity. Use existing SEO capability for foundational Search and access work where it fits; bring in specialist diagnosis when answer-led category, comparison, or commercial interpretation is the unresolved problem.”

That response may be less dramatic. It is more useful because the buyer can tell what information changes the recommendation.

Do not turn conditionality into an AI trick

Clear branch points are good commercial communication before they are a GEO tactic. They can help a buyer compare routes and avoid procuring a solution to the wrong problem. They may also give answer-led systems more explicit material for a bounded explanation under varying conditions.

None of this creates deterministic control.

For Google's AI features in Search, Google's official guidance says ordinary Search fundamentals remain relevant and that no new machine-readable files, AI text files, special markup, or special schema.org data are required.[1] llms.txt, arbitrary chunking, and over-focused structured data should not be sold as required switches for Google AI visibility.

The same restraint applies elsewhere. Publishing decision variables cannot guarantee inclusion, ranking, citation, recommendation, buyer action, attribution, conversion, or revenue. It improves the public account the business can control; it does not control the answer.

Make the next question easier to ask

Take one question your market asks in superlatives:

  • Which provider is best?
  • Which platform should we buy?
  • Should we hire a specialist?
  • Is this approach right for our company?

Do not answer it with a longer claim about yourself.

List the two or three missing facts that would genuinely change your recommendation. Put those branch points into the offer page, comparison material, methodology, or buyer guidance in plain language. Show when your offer fits, when another route is sufficient, and when the buyer should gather more information before choosing.

Then test whether the shortest responsible answer can preserve that conditionality.

The goal is not to make the company harder to recommend. It is to make the recommendation harder to fake.

A confident answer to an incomplete question is not decision support. Sometimes the most commercially useful answer begins by refusing to pretend the question is complete.

Sometimes the best AI answer is: it depends—here is what it depends on.

Source

[1] https://developers.google.com/search/docs/appearance/ai-features — Google Search Central: AI features and your website