Solutions by need 5 of 8

AI prompt tracking: see how answers to your buyers change

The same question gets a different answer from week to week. MentionX re-runs your buyer prompts on every engine, keeps every answer, and shows which ones gained you, dropped you or changed.

Start free See how it's calculated

Answers re-run this week800

Answers that changed42%

Now include you6%

Dropped you5%

Example: Vela CRM is an invented brand, scanned on Nova (200 prompts × 4 engines).

The definition

What is AI prompt tracking?

AI prompt tracking re-runs the same buyer prompts on AI engines over time and records how each answer changes: which brands it names, how it ranks them and which sources it cites. MentionX keeps every answer, so you can compare any scan with an earlier one.

Stable prompts: The same wording every week, so a different answer means the engine changed, not the question.
History per prompt: Each prompt keeps its answers by engine and date.
Up to 5 engines: One prompt, every engine in your plan, side by side.

What's the best CRM for a 20-person B2B sales team?

Most reviews point to Kestrel^#1 and Pipewise^#2 for a team of this size.

No sources cited in this answer.

Recorded for Vela CRMAbsentWeek 6 on ChatGPT: not named.

NamedRecommended

Example answers from one engine over three weeks. Invented brands.

Four kinds of change

Every re-run answer moved or held

MentionX compares each answer with the same prompt and engine last week. The changes are counted, never estimated.

Unchanged

Same brands, same order, same verdict for you.

…Kestrel and Pipewise, as last week.

Dropped you

You were named last week and aren't now.

Week 7: …Vela CRM. Week 8: ~~Vela CRM~~

Changed, you held

The wording, ranking or sources moved; your verdict held.

Kestrel moved from #2 to #1…

Now include you

You're named this week and weren't last week.

Week 8: 2. Vela CRM, for fast setup…

Answers that changed3 kinds

Dropped you

Now include you

The calculation

How answer changes are counted

Each square below is one prompt on one engine, compared with the same prompt and engine last week. Pick a count to light up the answers it includes.

Now include youDropped youChanged, you heldUnchanged

38%

ChatGPT

41%

Gemini

44%

Google AI Overviews

45%

Google AI Mode

Each block is one engine: 200 prompts, 200 squares.800 answers this scan

Answers that changed = answers that changed since last weekall answers

333 ÷ 800 = 42%800 = 200 prompts × 4 engines

Net this week: +4 answers name Vela CRM compared with last week.

Example data for an invented brand. Counts are calculated in code from stored answers; formulas are published on the methodology page.

Answers that changedAnswers that dropped you

Week over week

Some weeks, everything moves

Engines update their models and sources. Tracking the same prompts shows when a whole category shifts, and whether you shifted with it.

Answers that changedWeek 1 to week 8+11 pts

Answers that dropped youWeek 1 to week 8−1 pts

Example data for an invented brand.

The method

How to track AI prompts, step by step

Day 0
Build the prompt set
MentionX generates buyer prompts from your URL and market across six intents. Edit, add or import your own.
Day 1
Lock the wording
Keep prompts stable, so changes in the answer aren't changes in the question.
Week 1
Read the baseline
Every answer, by engine, with the brands it names and the sources it cites.
Weeks 2–4
Watch the movers
Each week, sort prompts by what changed: gained you, dropped you or shifted.
Monthly
Retire and add
Swap prompts that no longer matter for new ones; history stays with each prompt.

Reading the results

Five patterns worth acting on

Most answers hold from week to week. The ones that move are where to look.

Every chart here uses one colour code

Coral: absent, or a rival aheadBlue: mentioned, not pickedMint: recommendedSilver: first choice, or you

Many answers change in one week

An engine updated its model or sources.

First move: Check whether your verdicts moved with the category.

Weekly Monitoring

A prompt drops you on one engine

That engine replaced a source that named you.

First move: Open last week's answer and compare the citations.

Answers & Evidence

You climb on a prompt but not to #1

You're gaining trust, but the top pick hasn't changed.

First move: Read the reasons given for the #1 pick.

Competitor Tracking

Brand prompts hold, category prompts swing

Discovery answers are still forming in your category.

First move: Put category prompts first in your content plan.

Intent Coverage

A prompt never names a tracked brand

Engines don't answer it with brands, so it measures little.

First move: Rewrite it the way a buyer would ask, or retire it.

Prompt Library

The tools behind it

The MentionX features that do the work

Plans

How much each plan tracks

Answers per scan is prompts × engines, re-run 4 times a month. Spark and above include unlimited seats and competitors; Starter includes 3 of each.

SparkSelf-serve

$149 /mo

100 prompts × 3 engines = 300 answers a scan

Each square is 10 answers1,200 a month

ChatGPT, Gemini and Google AI Overviews

Choose Spark

NovaSelf-serve

$249 /mo

200 prompts × 4 engines = 800 answers a scan

Each square is 10 answers3,200 a month

Adds Google AI Mode

Choose Nova

GalaxySelf-serve

$599 /mo

300 prompts × 5 engines = 1,500 answers a scan

Each square is 10 answers6,000 a month

Adds Perplexity

Choose Galaxy

100 prompts per scan on Spark, 200 on Nova and Galaxy.

Questions

Prompt tracking, answered

The questions buyers ask before they start. Your first scan answers the rest with your own data.

Read the methodology

What prompts should I track for AI visibility?

The questions buyers ask AI across the journey: category, use cases, buying criteria, comparisons, alternatives and your brand. MentionX generates them for you.

Can I add my own prompts?

Yes. Edit any prompt or paste a list, and MentionX sorts them into the six buyer intents.

Does MentionX show prompt search volume?

No. AI engines don't publish how often prompts are asked, so MentionX doesn't invent volumes.

How many prompts can I track?

100 per scan on Spark and 200 on Nova and Galaxy, on every engine in your plan.

Why do AI answers change from week to week?

Engines update their models, indexes and sources, and competitors publish new pages. Re-running the same prompts shows which changes affect you.

Can I compare this week with an earlier scan?

Yes. Every stored answer is kept with its prompt, engine and date, so scans can be compared side by side.

Other needs

Start from a different question

Free forever, no card needed. Free is a plan ($0, no card, ChatGPT): 20 prompts the first week, then 10 a week. There is no 7-day trial

See how AI answers your buyers this week

Prefer to talk first?

Talk to the founder· 20 min

Your first scan stores every answer to your prompts, ready to compare with next week's.