MayaBook a demo
Back to Blog
AI VisibilityAgenciesNovember 19, 2026

AI Visibility Snapshot for a Sales Pitch: What to Show a Prospect and What Not to Promise

What an agency should show a prospect in an AI visibility snapshot before signing: buyer questions by engine, competitors and cited sources from the same answers, and what the snapshot cannot prove.

Prefer Maya AI in Google

Highlight our stories in Search, AI Mode & AI Overviews.

An AI visibility snapshot for a sales pitch should show a prospect a few of its buyer questions on several AI engines over a few days: whether the brand is named, which competitors appear in the same answers, and which sources those answers cite. The snapshot opens the GEO conversation rather than forecasting results. The method is the same in any global market, and Maya's agency workflow starts with this step. The exception is a prospect who wants traffic proof, which needs analytics access granted only after signing.

Below you will find which evidence to bring to the first meeting, what goes into the snapshot, what it cannot prove, how to present it, and how to turn each finding into a test your proposal can name. If the prospect has already signed and you need the recurring client report, the monthly AI visibility report template for agencies is the closer fit.

Start from what you can see before the prospect signs

Before a contract, an agency usually has no access to the prospect's Google Analytics or Search Console. Traffic, conversions and query data are out of reach, so the only evidence you can bring is what AI assistants say when a buyer asks about the prospect's category. A pre-sales snapshot is built entirely from that answer side.

Many pitches still open with a screenshot of one ChatGPT answer. A single answer is a weak opening for two reasons: AI answers change between runs, and a signed-in account can personalize what the seller sees. Maya's guide for reporting AI visibility to a CMO explains both effects, and the seven-metrics guide explains why one run is a sample rather than a measurement. A marketing manager who knows either point has good reason to discount a screenshot.

Choose the evidence before the meeting

Sellers often treat three different checks as one "free AI audit": a manual question in a chat window, a technical audit of one page, and a snapshot of AI answers. They answer different questions, and only one of them shows where the prospect stands against competitors. The table puts them next to the tracking that follows a signed contract.

EvidenceWhat it shows the prospectCompetitors and cited sourcesWhat it cannot prove
A manual check in a chat windowWhat one assistant said once, in one sessionOnly the names and links in that single answerWhat other buyers see, or any trend
A technical page audit (Maya's free LLM Visibility Score)Access, delivery and content-quality checks for one page, with a 0–100 score and prioritized fixesNot shown, because the audit inspects the submitted page rather than AI answersThe brand's real mention, citation or recommendation rate
A pre-sales answer snapshotWhether the prospect is named on its own buyer questions, engine by engine, and how it is describedCompetitors named in the same answers and the sources those answers citeA stable rate, a trend, or the result of future work
Ongoing tracking after signingMention rate, share of voice, citation share, recommendation rate and sentiment as averages and trendsCompetitors and cited sources on every tracked question, over timeThat a single change caused a movement, without dated changes and repeated readings

The technical audit and the answer snapshot work well together, but they should never share a slide title. Maya's audit page states that its score "is not your brand's real mention, citation, or recommendation rate in AI answers" (Maya LLM Visibility Score). Present the audit as "can AI crawlers read your key page" and the snapshot as "where you stand in the answers".

What goes into an AI visibility snapshot for a prospect

A snapshot answers one question for the prospect: when a buyer asks an AI assistant about this category, what happens to our brand? Four parts are enough to answer it, and each part maps to a decision the prospect will later make with you.

Pick a few buyer questions from the prospect's sales conversations

Choose a handful of questions a real buyer would ask before contacting the prospect: a category recommendation, a comparison, and a question about fit for a specific customer type. Avoid questions that already contain the prospect's name, because those only test whether the assistant knows the brand, not whether it recommends it. Maya's methodology makes the same distinction: "What does Maya AI do?" and "Which tools should I consider for measuring AI visibility?" are different questions (Maya methodology). The guide to building a prompt set covers question selection in depth; for a pitch, a short list drawn from the prospect's own sales conversations is enough.

Break the answers down by engine

Show the result per engine rather than as one blended figure. A prospect that is missing from ChatGPT answers but named on Perplexity has a different problem from one that is missing everywhere. Maya tracks ChatGPT, Gemini, Perplexity and Google AI Overview daily on every client, with Claude, Grok, Copilot, Google AI Mode and DeepSeek available per client (Maya for agencies).

Show competitors and cited sources from the same answers

Competitor names carry weight only when they come from the same answers as the prospect's result. Maya's agency workflow describes the pre-sales step as running "a visibility snapshot for a prospect" and walking into the pitch "with their AI share of voice against three competitors" (Maya for agencies). Use the competitors the prospect names in the first call, not the largest companies in the category.

Then list the sources the answers cite, grouped by type: the prospect's own pages, competitor pages, third-party comparison articles, review sites and forums. A source list turns "you are not in the answer" into "the answer is built from these pages", which is where a GEO proposal can start.

Write the unit and the dates on every number

Write the collection dates, the engines and the counting unit next to each figure. Maya's methodology gives an example: when a question runs on four platforms and the brand appears in one answer, the response-level rate is one out of four, while an evaluation-level measure can record that same evaluation as containing a mention (Maya methodology). Two honest figures can look contradictory if the unit is missing, and a prospect who spots that gap will question every other number on the slide.

Count fresh answers, not days. Maya's methodology states that on projects without uncached collection a platform answer may be reused for a short window, three days by default, and that every reused answer is flagged on its record (Maya methodology). Leave reused answers out of the snapshot's count, or set the prospect's project to uncached collection.

Say what the snapshot cannot prove

A snapshot is a few samples, not a measurement. Maya's methodology states that each tracked question is collected once a day on each platform and that "a single answer should not be generalized to an entire market" (Maya methodology). Over a few days, each question collects only a few answers per engine. Maya's seven-metrics guide calls one pass "a sample, not a measurement" and reads stability only from repeated runs (Maya seven-metrics guide). A few days of answers can show direction and examples, not a stable rate.

Nor can a snapshot promise a place in future answers. Google states for its own AI features that "indexing and serving isn't guaranteed" (Google Search Central), and its guidance on hiring an SEO warns that "no one can guarantee a #1 ranking on Google" (Google Search Central). A promised AI ranking contradicts both statements, and a prospect who has read them may count it against the agency. Put one sentence on the snapshot slide that says what the data covers and what it does not.

How agencies pitch AI visibility to a prospect in the first meeting

Present the snapshot in the order the prospect experiences the problem. Start with one buyer question and the answers it received on each engine, then show which competitors those answers named and which sources they cited. Only after that, show what you would test first and how you would measure it. Ending on the measurement plan keeps the meeting on a scope the prospect can approve, rather than on a number you cannot promise.

The same questions come up in most first meetings. The table pairs each situation with what to show and what to avoid saying.

Situation in the meetingWhat to showWhat to avoid
The prospect's team already ran a free AI visibility checker and asks how accurate it is compared to manual testingThe engines, dates and number of answers behind your snapshot, next to the single answer their manual test producedCalling either result a visibility rate
The marketing manager asks you to audit brand mentions across AI platforms before signingThe engine-by-engine table from the snapshot, with the collection datesBlending the engines into one score
The prospect wants a report on why the brand is missing from AI answersThe cited sources grouped by type, with a stated hypothesis for each patternStating a cause as proven
The prospect asks which metrics a free visibility audit includesThe technical audit's own categories, labelled as page checksPresenting the audit score as the share of voice
Finance asks how the agency would measure ROI from AI search visibility workThe tracking plan after signing, with analytics access addedProjecting revenue from the snapshot
The prospect asks whether you can guarantee a place in ChatGPT answersGoogle's own statement that serving is not guaranteed, and what you will measure insteadAny guarantee

The last two rows are where a pitch most easily overpromises. Answer them with the measurement plan, not with a forecast.

Turn each finding into a test the proposal can name

The snapshot earns the second meeting only if each finding leads to a concrete change the prospect can approve. Write the proposal as a short list of tests, not as an outcome. Each test names the page or source to change, the hypothesis behind it, and the questions that will be read again after the change.

Pair each finding with a hypothesis, not a cause

An observation is what the snapshot shows: "the answers cited a third-party comparison article and a competitor's pricing page, not your site". A possible cause is a hypothesis: "one explanation is that your pricing page does not state who the product is for". Keep the two apart on the slide. A research finding about what helps pages get cited is worth testing, but it does not prove the cause on the prospect's own page.

Propose one or two changes as the first tests

Limit the first scope to one or two changes the prospect's team can make quickly, such as stating plan limits and customer fit on the pricing page, or correcting an outdated description in a third-party article. A short scope makes the result easier to read and keeps the proposal free of promises.

Promise the measurement, not the result

After signing, the same snapshot questions move into tracking on the same engines, and the date of each change is recorded. The proposal can commit to that reading schedule and to reporting what changed, without committing to a ranking. The seven-metrics guide explains how many runs make a reading stable.

Illustrative example: one buyer question in a pitch

The following example is illustrative and does not describe a real prospect. An agency prepares a pitch for an invoicing software company and tracks the question "which invoicing tool suits a freelance designer" on several engines over several days, counting only freshly collected answers.

Observation. Across engines and days, the answers name two competitors and rarely the prospect. The cited sources are a third-party comparison article and a competitor's pricing page; the prospect's own pages are not among them.

Possible cause. One explanation is that the prospect's site does not say which customer types its plans suit, which is the detail the answers repeat. Another is that the comparison article describes an older version of the product.

What Maya shows. Maya's Competitive Benchmarking compares the prospect with competitors on the same monitored questions, and Source Analysis lists the domains and pages the answers cite.

Action and check. The proposal names two tests: add customer-fit and plan-limit statements to the pricing page, and ask the comparison article's publisher to update its description. After signing, the agency tracks the same question on the same engines, marks both change dates, and reports whether the prospect's pages appear among the cited sources.

Limits of this guide

The guide does not set prices for GEO services, estimate the return on GEO work, or compare AI visibility tools; the guide to choosing AI visibility tools for agencies covers tool evaluation. Metric definitions are in the seven-metrics guide, and recurring client reporting after signing is in the agency report template.

Frequently asked questions

How accurate are free AI visibility checkers compared to manual testing?

Both read a small number of AI answers, so neither produces a stable rate. A checker that records several engines and freshly collected answers covers more conditions than one manual question in a signed-in account, but AI answers vary between runs, and a reliable reading needs repeated runs on a fixed question set.

What metrics are included in Maya's free visibility audit?

Maya's free LLM Visibility Score returns an overall score from 0 to 100, category breakdowns for Access, Delivery and Quality, pass or fail checks with severity levels, an AI bot access table and recommendations prioritized by impact and effort. The page offers one free audit per day.

How can agencies measure ROI from AI search visibility monitoring?

Return on GEO work can only be read after signing, once the agency has the client's analytics. Google Analytics groups visits from sources such as ChatGPT, Gemini and Copilot in an AI Assistant channel, which can sit next to the visibility trend for the same period without being read as proof of cause.

Conclusion: Sell the measurement plan, not a ranking

A credible pre-sales snapshot shows a prospect a few of its own buyer questions, engine by engine, with the competitors and sources from the same answers and the unit written next to every number. The snapshot opens the conversation, the proposal names one or two tests, and tracking after signing shows whether they worked.

Maya's agency workflow starts with this step: an agency runs a visibility snapshot for a prospect and brings the prospect's AI share of voice against three competitors to the pitch, then tracks the same questions daily once the prospect becomes a client. Agency plans are quoted per roster on the agencies page, and the pricing page lists a 3-day free trial on its Starter, Premium and Enterprise plans.

Sources

About the author

Sanzhar Tuibekov

GEO researcher at Maya. Works on brand representation in AI answers and how agencies can scale AI visibility for their clients.

Ready to improve your AI visibility?

See how your brand appears across ChatGPT, Claude, Gemini, and other AI assistants.