No. It shows how AI engines answered a prompt for you, not how often anyone asks it, and not that inventory exists for it. Every card is inferred: treat it as a test worth running, not as confirmed demand.

Do I need to be spending already?

No. Three of the five types read only from your tracked answers, so the list works before any campaign exists. The two paid types need a campaign that has spent at least 50 in your account’s currency. Below that, there isn’t enough to judge.

What is the score out of 100?

A sort order. It weights your organic visibility, competitor presence and, when there’s spend, what the ads account reports. The weights are our hypothesis, not a measurement, and every card shows the evidence underneath so you can check the ranking yourself.

Where does the evidence come from?

Each opportunity lists its rows with a metric, a baseline, a current value, a period and a source: either your connected ads account or LLMagnet’s own tracked answers. The prompt is named, so you can open the answers it came from.

Advertise

Where paid could help,
and where it can’t

LLMagnet reads the answers it already tracks for you and lists the questions where paid might be worth a test, with the evidence behind each one and a label saying how far that evidence goes.

Book a demo See what it’s inferred from

30-day free trial · no credit card

Ad opportunities for Hardly: questions where paid advertising could help, read from the AI answers already tracked

The problem

The worst ad budget is the one buying something you already had

A ChatGPT Ads account can tell you what a campaign spent. It can’t tell you that the question you just bought was one AI already named you for, unpaid, in six answers out of seven. That’s money spent to stand where you were already standing, and there’s no line item for it.

Keyword plannersquat proof leggings

KeywordAvg. monthly searches

squat proof leggings—

high waisted gym leggings—

best leggings for squats—

Not available for chat answers

Lululemon namedHardly —

Visibility gap“best squat proof leggings”

You0 of 7

Score80

Inferred · high confidence

There’s no keyword tool for this

The planners you’ve used for twenty years rest on counted searches. Nobody counts the questions inside a chat for you, so the starting point has to be something you can actually see.

Search volume, counted by the platformAnswers, run and stored by us on a schedule

Keyword, match type, bidA question, and whether you were named in the answer

“Is there demand?”“Is there absence?” You can see that one.

A signal with the platform’s authorityA signal with its confidence label attached

A question you lost, with the answers that prove it

Each opportunity opens onto the prompt it came from, how often you were named, how often a competitor was, and a next step you can read before you spend anything. Hover the numbers; open the answers.

Visibility gapHigh priority“best squat proof leggings”1

Visibility gap score80High confidence

You appear in 0 of 7 tracked answers to this question2

You appear in0%

Competitor presence3100%Lululemon · 7 of 7

Suggested next step4Draft a paused ad group with a small test budget and conversion tracking, then review it before anything goes live.

Inferred from last 7 days of tracked answers.5

The seven answers“best squat proof leggings” · Oct 26 – Nov 1

For squats, most lifters point to Lululemon’s Align: opaque at full depth, and high enough to stay put.

Lululemon namedHardly not named

On the screen, in the product

These are inferred opportunities. A tracked prompt shows how an AI engine answered it for you. It does not prove how often anyone asks it inside ChatGPT, that ad inventory exists for it, or that an ad can be aimed at that exact question. Treat each one as a test worth running, not as confirmed demand.

Five types

Five things worth knowing before you spend

Every opportunity is one of five types, and the type tells you what kind of decision you’re being handed.

01Visibility gap
A question you’re tracking where you’re barely named, or not named at all.
Your tracked answers
02Competitor dominance
A rival is named across the answers to a question, and you aren’t.
Your tracked answers
03Paid performance gap
A campaign has spent real money and the account reports zero conversions.
Your ads account
04Organic strength
You’re already named here, for free. Paying for it buys nothing new.
Your tracked answers
05Unmeasured spend
A campaign is spending and nothing is measured, so there’s nothing to judge it by.
Your ads account

Each one carries a score from 0 to 100, a priority, a confidence level, the prompt it came from, your organic visibility, the competitor presence, and evidence rows naming where every number came from.

We won’t judge a campaign that has barely run

Neither of the two spend-based types appears until a campaign has spent at least 50 in your account’s currency. Below that there isn’t enough to say anything, and a confident verdict drawn from a handful of impressions is worse than no verdict at all.

£50

Rep Seamless · Leggings£0.00No verdict yet
Holiday prep£61.67Unmeasured spend
Shorts · Prospecting£209.28Unmeasured spend
Spring training · Brand£371.05Unmeasured spend

Organic strength

Sometimes the finding
is: don’t buy this

Organic strength is a type of its own. It marks the questions where AI already names you without a budget, so if a campaign is pointed at one of them, you’re paying for an answer you were getting anyway.

It’s the least exciting card in the list and the most useful one. Paid and organic teams rarely share a screen, which is how a brand ends up bidding against its own best answer. Here they’re two columns of the same card.

Organic strengthMedium priority

“best gym wear brands uk”

You appear in86%6 of 7 answers

Competitor presence29%Nike · 2 of 7

You’re named here without a budget. Check that no campaign is paying to stand where you already are.

Inferred from last 7 days of tracked answers · high confidence

Found it. Now fix it.

This page shows where paid might help. The optimization agent turns a decision into a prepared change, and waits for your approval.

Optimization agent

From a card to a change

The proposal carries the same prompt, its evidence rows, a confidence label and a date it expires. Dismiss weighs as much as Approve.

Optimization agentExpires Nov 8

From: Visibility gap · “best squat proof leggings”

Create a paused test campaign, £10.00/day

Named in answers0 of 7LLMagnet

Lululemon named7 of 7LLMagnet

Spend on this topic£0.00ChatGPT Ads

Medium confidenceCreated paused · you activate it

DismissApprove change

What it’s allowed to do

Four changes, and no others

Pause a campaign, activate one, move a budget, or create one. It reads the campaign back from your account before it reports anything as applied.

Pause a campaign
Activate a campaign
Move a budget
Create a campaign

Meet the agent

The score is a sort order, not a measurement

It ranks the list. It doesn’t estimate return, volume or cost.

From our own code, published as it is

Weights are a product hypothesis, not a measurement.

What it’s built from. Your organic visibility on the prompt, how present a competitor is in the same answers and, once you have spend, what the ads account reports. Those inputs are weighted, and we chose the weights.
What the confidence label is for. Low, medium or high, moving with how much the underlying answers agreed. One answer and a pattern held across a week aren’t the same finding, and they aren’t labeled as one.
What you can do instead of trusting it. Open the evidence. Every card lists its rows and the prompt behind them, so the order can be checked rather than taken on faith. How we measure

EvidenceOrganic strength · “best gym wear brands uk” · 45

Metric	Baseline	Current	Period	Source
Named in answers	5 of 7	6 of 7	7 days	LLMagnet
Nike named	3 of 7	2 of 7	7 days	LLMagnet
Spend · Spring training	£318.40	£371.05	7 days	ChatGPT Ads
Conversions	Not measured	Not measured	7 days	ChatGPT Ads

1.7%of the time, four AI models cited the same domain for the same question

Answers move between days, too: run again two days later, Claude kept only 30% of the URLs it had cited, and Perplexity, the steadiest, 65%. A score read off one day’s answers would have no shelf life, which is why every card states its window.

Bill Widmer, Orbit Media, Sep 2026 · 72 questions, a directional sample

Why every card says “inferred”. We haven’t seen evidence, from anyone, that a gap in AI visibility predicts how an ad will perform. So we don’t imply it. The word stays on the card, and the evidence rows do the rest.

So this is keyword research for ChatGPT?

No, and we’d rather lose the comparison than make it. Keyword research starts from counted demand. This starts from answers we ran and stored for you, which proves absence, not appetite. It can’t tell you how often anyone asks the question, whether inventory exists for it, or that an ad can be aimed at it. Each card is a test worth running, and that’s the whole claim.

Opportunities are where you decide

What it proposes from theseOptimization agent

Create a paused test campaignMedium confidence · expires Nov 8

DismissApprove change

The account these are compared againstChatGPT Ads

£642.00Spend · Oct 26 – Nov 1 · Hardly UK

Cost per conversion · not measured

AI Visibility · the organic half every opportunity is read from Integrations · the ads account, and the store or CMS the agent writes to

Frequently asked questions

Want to see the product before you start? Take the product tour, or start a 30-day free trial.

No. An opportunity names the question an answer came from; it doesn’t mean an ad can be aimed at that question. Ad groups take context hints, which are hints about where an ad fits and guarantee nothing about any particular question.

See where you’re absent
before you buy it.

The list reads from tracked answers, so it has something to say before you’ve spent anything.

Book a demo