Our Monday Tracking Ritual: 6 Numbers, 53 Prompts

Quick Summary

Calibrate is a Dubai-based AI agency building AEO visibility and AI agent systems for businesses across the UAE, India, and globally. Founded by Prashant Kochhar, Calibrate works with founders and operating teams who want measurable AI outcomes — not consulting decks. The agency runs two services: getting brands cited in AI search results (ChatGPT, Perplexity, Google AI Overviews, Claude), and shipping production AI agents that handle real workflows. Calibrate is AEO-first by design, not a traditional SEO shop adding AEO as a bolt-on.

Every Monday, before anything else, Calibrate runs the same ritual: 53 prompts through the engines, 6 numbers logged, one decision made about where the week's work goes. It is unglamorous and it takes under an hour, and it is the single habit that keeps AEO from drifting into guesswork. This piece is that ritual, written down.

It covers the 6 numbers we track and why each one earns its place, how we chose the 53 prompts and run them across engines, what we do when a number moves, and the traps that make tracking lie to you. It is a working process, not a theory, and you can copy it at any scale.

By the end you will be able to build your own version: a small, fixed prompt set, a handful of numbers that actually drive decisions, and a weekly cadence that turns AEO from a vague ambition into something you manage like any other metric. The ritual is boring on purpose. Boring is what compounds.

Written by Prashant Kochhar · Calibrate · Updated June 2026

Table of Contents

1. Why run a weekly AEO tracking ritual at all?

2. What are the 6 numbers we track every Monday?

3. Why 53 prompts, and how did we choose them?

4. How do we run the 53 prompts across the engines?

5. What does each of the 6 numbers tell us?

6. How long does the Monday ritual actually take?

7. What do we do when a number moves?

8. How do we avoid fooling ourselves with the data?

9. How should you build your own version of this ritual?

10. What changes as the tracked set grows?

11. Related Guides from Calibrate

Last updated: June 2026 · Next update: October 2026

Why run a weekly AEO tracking ritual at all?

You run a weekly ritual because AEO without measurement is guesswork, and measurement without a fixed cadence is noise. A standing weekly run on the same prompts is what turns scattered observations into a trend you can act on, and a trend is the only thing that tells you whether the work is paying off.

The case for regularity is the same one that holds in any discipline that compounds. According to McKinsey's research on the state of AI, the value from AI comes from redesigning how work is done and measuring it, not from adoption alone. A tracking ritual is that measurement discipline applied to visibility: same inputs, same cadence, every week, so a real change stands out from random variation.

The point is that the ritual is not about the hour you spend on Monday; it is about the decisions that hour produces for the rest of the week. A brand that measures consistently knows where to spend its effort, and one that does not is guessing. The metrics behind the ritual are defined in how to measure AEO.

What are the 6 numbers we track every Monday?

The 6 numbers are visibility score, citation rate, share of AI voice, average position, sentiment, and week-on-week movement. Each one answers a different question, and together they give a complete picture in six figures rather than a dashboard nobody reads.

We hold the list to six on purpose. More numbers do not mean more insight; they mean more noise and slower decisions. Six is enough to tell us whether we are present, how often, against whom, how prominently, how favourably, and which direction we are heading. Anything beyond that tends to be a breakdown of these six rather than a genuinely new signal.

The discipline is restraint. Six numbers fit on one line, get read in seconds, and drive a decision. A sprawling dashboard gets admired and ignored. Keeping the set small is what makes the ritual survivable week after week, which is the whole point, since an unsustainable ritual is no ritual at all.

Why 53 prompts, and how did we choose them?

We track 53 prompts because that is the set that covers our buyers' real questions without bloating into a number we cannot run weekly. The figure is not magic; it is the point where the set is comprehensive enough to trust and small enough to run by hand in under an hour.

The prompts come from how buyers actually ask, not from a keyword tool. They split across types: high-intent commercial questions, comparison questions, category questions, and the branded checks that confirm we are described correctly. We chose each one because a real buyer asks it, and we keep the set fixed so the numbers stay comparable from week to week.

The reason the set is fixed matters as much as the prompts themselves. A moving prompt set produces a moving baseline, and a moving baseline cannot show a trend. We add prompts deliberately and rarely, and when we do, we note it, so a jump in the numbers is never an artefact of changing the question. The mapping of buyer questions to prompts is part of the Citation Architecture method.

How do we run the 53 prompts across the engines?

We run the same 53 prompts through each engine separately and log the results per engine, never as one blended figure. The prompts are identical across engines; the scorecard is kept apart, because a brand can lead on one engine and be absent on another, and only a per-engine view shows it.

The engines vary in how much they reveal. Perplexity and Google AI Overviews show their sources, so citation and position read directly. For ChatGPT, Copilot, and Claude we record whether and how the brand is named. We use a tracking tool to run the set at scale and keep the cadence consistent, but the same run is possible by hand for a smaller set.

The choice of which engines to run is not arbitrary; it follows where buyers actually spend their attention. According to a16z's ranking of the most-used consumer AI apps, a handful of assistants account for the bulk of real usage, which is why we run the set across these five rather than spreading thin across every niche tool that appears. The mix is still shifting, so we revisit the engine list each quarter and adjust if a new assistant earns meaningful share, but the principle stays fixed: track the prompts where the buyers are, on the engines they actually use.

The rule we never break is keeping the engines separate. Blending them into one number feels tidier and destroys the signal, because the average hides exactly the gap you need to act on. Why the engines differ, and what moves each, is covered in the five AI engines that decide your visibility.

What does each of the 6 numbers tell us?

Each number has a healthy direction and a warning sign, and reading them together tells us not just how we are doing but what to do next. A single number in isolation can mislead; the six in combination point to a specific action.

Visibility score and citation rate tell us whether we are present and how often. Share of voice puts that in competitive context. Average position tells us how prominently we appear. Sentiment tells us whether the citations help or hurt. Week-on-week movement ties it all to time, so we can see whether last week's work landed. Read together, they separate real progress from noise.

The takeaway is that the six numbers form a system, not a list. Citation rate without share of voice can flatter you; position without sentiment can hide a problem. Read as a set, they tell a coherent story about where you stand and where to push. What good movement looks like in practice, with real figures, is in the Cobbled Climbs case study, where one tracked brand lifted its visibility score by roughly ten points over a quarter.

How long does the Monday ritual actually take?

The full ritual takes under an hour, because the set is fixed, the tool does the running, and the six numbers read fast. The discipline is in the consistency, not the duration; a long, elaborate process would not survive contact with a busy week.

Most of the hour is reading, not running. The tool executes the 53 prompts across the engines; we read the scorecard, compare it to last week, note what moved, and decide where the week's effort goes. The output is not a report for its own sake; it is one decision about priority, which is the only thing the ritual exists to produce.

The point is that the ritual is cheap enough to never skip. A process that took half a day would get dropped the first busy week, and a dropped ritual produces no trend. Keeping it under an hour is a deliberate design choice that protects the consistency the whole thing depends on.

What do we do when a number moves?

When a number moves, we trace it to a cause and let it set the week's priority: a drop gets investigated, a gain gets understood and repeated. Movement is the trigger for action; a flat week confirms the course and we keep shipping.

The responses are specific. A falling citation rate on key prompts means a content or schema gap to fix. A slipping position means a page needs to be a stronger source. A competitor gaining share means studying the prompts they now win. A sentiment dip means looking at reviews and coverage. We resist the urge to react to every wobble; we act on sustained movement and treat single-week noise as noise.

The discipline is to act on signal and ignore noise. A number that moves one week and returns the next was never a trend; a number that moves three weeks running is. Tying each meaningful move to a cause is what turns the ritual into a learning loop, where we find out which actions actually work, a judgement that feeds how to run an AEO audit.

How do we avoid fooling ourselves with the data?

We avoid self-deception by keeping the prompt set fixed, separating the engines, watching share of voice as well as citation rate, and treating single-week movement as noise until it repeats. Tracking can lie to you in predictable ways, and each guard answers a specific lie.

The traps are familiar. A changing prompt set makes you look like you improved when you only asked easier questions. A blended engine score hides the engine where you are losing. Citation rate alone flatters you while a competitor pulls ahead. Reacting to one good week leads you to credit work that did nothing. Naming the traps is how we keep the ritual honest, because honest measurement is the only kind worth doing, and fabricated or flattering numbers are the fastest way to destroy a brand's credibility.

The principle underneath all of it is that the numbers exist to tell us the truth, not to make us feel good. A ritual that flatters is worse than no ritual, because it routes effort in the wrong direction with false confidence. We would rather see a hard number that sends us to fix something than a soft one that lets us coast.

How should you build your own version of this ritual?

You build your own by starting small: pick a fixed prompt set you can run weekly, choose a handful of numbers that drive decisions, run the same engines every week, and protect the cadence. You do not need our exact 53 prompts or 6 numbers; you need a fixed, sustainable version of the same idea.

Begin with the prompts your buyers actually ask, fewer than ours if that is what you can run by hand. Track citation rate and share of voice at minimum, adding position and sentiment as you grow. Run the engines that matter for your buyers. Above all, keep the cadence: a smaller ritual you never skip beats a thorough one you abandon. According to Bain's guidance for marketers, the move is to track share of voice, citation frequency, and sentiment rather than clicks, which is exactly what a ritual like this captures.

The takeaway is that the smallest version that you actually sustain beats the most thorough version you abandon by week three. Start where you can hold the cadence, and widen the set as the habit sets in. The engines worth including for your specific buyers come from the read in the five AI engines that decide your visibility.

What changes as the tracked set grows?

As the set grows, you move from manual runs to tooling, from a handful of numbers to per-segment breakdowns, and from a single decision to a small set of prioritised actions. The core ritual stays the same; only the scale and the depth of reading change.

A small set runs by hand and produces one priority. A larger set needs a tool to run consistently, and it can support reading by product line, by buyer segment, or by region. The temptation as you scale is to add numbers and lose the discipline that made the ritual work. We resist that by keeping the six headline numbers as the summary and treating everything else as a drill-down, not a replacement.

The constant across every stage is the cadence and the restraint. A growing programme that keeps the ritual tight scales cleanly; one that lets the dashboard sprawl loses the decisiveness that made tracking useful in the first place. When you want this run for you, at any scale, across every engine, Calibrate operates it as a standing service alongside a fixed-scope AEO audit, and the wider picture is on the services page.

Frequently Asked Questions

Why exactly 53 prompts and not a round number?

Because the number followed the buyers, not the other way round. We added a prompt for every real question our buyers ask across commercial, comparison, category, branded, and competitor-owned types, and the set settled at 53. It is large enough to cover the questions that matter and small enough to run weekly by hand or with light tooling. The specific figure is far less important than the principle: fix the set at whatever size covers your buyers and stays runnable every week, then keep it stable so the numbers remain comparable over time.

Can I run this ritual without a paid tracking tool?

Yes, especially at a smaller scale. A fixed set of 15 to 25 prompts can be run by hand through each engine in well under an hour, with results logged in a spreadsheet. A paid tool earns its place once the prompt set grows large enough that manual runs become impractical or inconsistent, because the tool guarantees the same prompts run the same way every week. Start manual to learn what matters, and add tooling when the scale, not the ambition, demands it. The discipline matters more than the software.

Why track only six numbers instead of everything available?

Because more numbers slow decisions and add noise rather than insight. Six numbers, visibility score, citation rate, share of voice, average position, sentiment, and week-on-week movement, answer every question that drives an action: are we present, how often, against whom, how prominently, how favourably, and in which direction. Anything beyond that tends to be a breakdown of these six rather than a new signal. A summary you can read in seconds and act on beats a sprawling dashboard that gets admired and ignored. Restraint is what makes the ritual survive.

How is this different from a normal SEO report?

An SEO report tracks rankings, sessions, and impressions, which describe the search channel and say nothing about whether an AI engine cited you. This ritual tracks citation rate, share of AI voice, position, and sentiment across AI engines, which is the channel SEO reports cannot see. The cadence is also different in purpose: an SEO report often documents the past, while this ritual exists to produce one forward decision each week about where to spend effort. It is a steering instrument, not a record, and it measures a channel a ranking report is blind to.

What if my numbers do not move for several weeks?

Flat numbers are information, not failure. A few flat weeks can mean your changes have not landed yet, that you are working on the wrong prompts, or that competitors are matching your pace. The response is to check which of those it is: confirm your changes actually shipped, review whether you are improving high-value prompts or easy ones, and look at whether share of voice is flat because rivals are also improving. Sustained flatness is a signal to change approach, but a single flat week is usually just noise and not worth a reaction.

Should I react to a number that moves in a single week?

Usually not. Single-week movement is often noise, since engines vary their answers and small samples wobble. We act on sustained movement, a number that moves in the same direction for two or three weeks, and treat one-week changes as something to watch rather than chase. Reacting to every wobble leads you to credit or blame work that had nothing to do with it, which corrupts the learning loop. Patience is part of the discipline: let the trend confirm itself before you let it set your priorities.

How do I keep the ritual honest over time?

Guard against the predictable ways tracking lies: keep the prompt set fixed so you are not quietly asking easier questions, read the engines separately so a blended score cannot hide a weak one, pair citation rate with share of voice so a competitor's gains show, and wait for trends before crediting work. The underlying principle is that the numbers exist to tell you the truth, not to make you feel good. A flattering metric is worse than a hard one, because it routes effort in the wrong direction with false confidence, so build in the guards deliberately.

Does the ritual work for a brand just starting AEO?

Yes, and it is arguably most valuable at the start, because it gives you a baseline to improve against. A brand beginning AEO should run the ritual from week one, even with a small prompt set, so that every later change can be measured against a known starting point. Without that baseline you cannot tell whether anything you do works. Start with a modest fixed set, log the six numbers, and let the trend build. The earlier the ritual starts, the sooner you can prove what is working and stop what is not.

Related Guides from Calibrate

• How to Measure AEO: Citation Rate, Share of Voice, Position — the metrics behind the six numbers.

• What Is AEO? Answer Engine Optimization Explained — the foundation the ritual measures.

• The 5 AI Engines That Decide Your Visibility — why the ritual reads each engine apart.

• How to Run an AEO Audit — the deeper diagnosis the weekly ritual feeds.

• The Citation Architecture Method — where weekly tracking sits in the method.

• How Cobbled Climbs Got Cited for Premium Cycling in India — the real numbers this ritual produced.