Published on: November 3, 2026
As clients push to appear in ChatGPT, Gemini, Perplexity and Google’s AI answers, Mavlers Agency’s Darshan Modi argues that agencies should sell strategy, execution and measurable progress rather than placements, agree a baseline before committing to outcomes, and report plainly what can be measured, what can be inferred and what cannot be established.
Clients in the US, the UK, Australia and the Middle East are asking agencies the same question: will our brand show up in ChatGPT, Gemini, Perplexity and Google’s AI answers?
Some agencies are answering with a confidence the platforms do not support. AI responses vary, the systems behind them belong to someone else, and a client who is promised certainty will eventually ask for the evidence.
Here is how agencies should sell, deliver and report GEO so that the conversation still holds up when they do.
An agency can commit to improving a brand’s visibility, relevance and discoverability across AI platforms, through structured research, execution and measurement. That is work the agency controls, and it can be scoped, delivered and reviewed.
What an agency should not promise is guaranteed mentions, citations, recommendations or leads within a fixed timeline. Those outcomes depend on platforms the agency does not own, and a guarantee is only as good as the control behind it.
The principle is simple. Sell the strategy, the execution and the measurable progress. Do not promise control over platforms you do not run.
“Sell the strategy, the execution and the measurable progress. Do not promise control over platforms you do not run.”
A client report should be built around three areas:
Agencies should avoid presenting isolated screenshots, individual prompt positions or a single visibility score as proof of performance. Each of them shows one moment or one number, and none of them explains why it moved.
A citation is not necessarily a recommendation, and visibility on its own is not a business outcome. A report that treats them as the same thing will overstate the result, and the client will notice.
The hardest question a client asks is whether GEO actually brought leads. This is where agencies need to be most careful.
When qualified leads or conversions can be tracked from AI referrals, they should be reported. Many buyers do not click through, though. They see a brand in an AI answer and convert later through another channel, and at that point attribution gets difficult.
Every report should separate three things: what can be measured, what can reasonably be inferred and what cannot be established. Measured results sit in one place, inferences are labelled as inferences, and gaps are stated as gaps.
An agency should not take credit for leads it cannot support with evidence. Nor should it assume GEO has no value simply because direct referral traffic is limited.
One recurring risk is positioning GEO the way traditional SEO is positioned, with predictable rankings, consistent citations and guaranteed timelines. AI responses can differ by platform, location, session and context, and even identical prompts may not return the same mentions or citations.
An agency that promises consistent placements on the strength of a few successful prompt checks is setting up its own difficult conversation. Expectations harden quickly, and the next variation in an AI answer looks like a failure.
The approach is to settle the measurement methodology, the baseline, the limitations and the success criteria before committing to any outcome. That sequence gives both sides a fixed reference point, so later movement can be read against something agreed instead of against whichever prompts happen to look best that week.
For an agency starting GEO tomorrow, the advice is to begin with the client’s business problem. A package or a tool can come later. That means understanding what the client’s customers are asking, where competitors are appearing, how the brand is currently represented, and what can realistically be improved.
An agency should be able to explain what it is improving, why that matters, and what evidence supports each recommendation. If it cannot, the report is not ready to be shown.
The long-term value will sit with agencies that help clients make better decisions on credible evidence, and the size of their promises will matter much less.
