Track where AI recommends your business

What does Radar actually measure?

When someone in Toronto asks an assistant “best med spa for Botox in North York”, the answer names a short list of businesses and moves on. There is no ranking page, no impression count, and nothing in your analytics that tells you it happened. Radar is the instrument for that blind spot.

For every question, in every run, it records:

• Whether you were named — and, when you were, where in the answer you appeared. Being the first business listed is not the same result as being the fourth, and the two are stored differently.
• Which competitors were named instead. Every other business the answer mentioned is captured and resolved to a canonical entity, so “Glow Atelier”, “Glow Atelier Toronto” and “glowatelier.ca” count as one rival rather than three.
• Which sources the answer leaned on. The links and domains behind the answer, each classified by type — the business’s own site, a review site, a directory, a Google profile, a local publisher, a government page. This is the part that tells you what to go and fix.

All of it is tracked over time. A single snapshot of an AI answer is close to meaningless: run the same question twice in a week and the names can change. The measurement that matters is the trend across dozens of questions and many weeks, which is the only shape in which this data says anything reliable.

Where do the tracked questions come from?

Two sources, and the split is deliberate.

The base pack is a curated set of questions for your category and your market — not generic industry phrases, but the ones people in your neighbourhood actually ask, ranked by estimated demand. cite0 builds and runs these packs ahead of any individual client, which is why a new account has a competitive baseline on day one instead of after a month of collection. Starter tracks 100 base questions, Boost 200, Dominate 300.

Custom questions are yours: the phrasing specific to your business, your signature service, the objection you hear on every consult call. Starter includes 5, Boost 20, Dominate 50. They are built from templates with catalogue-sourced slots — a service, a competitor, a landmark, a concern — rather than free-typed, so a question can be validated, priced and compared against the others rather than quietly failing to be answerable.

The demand estimates behind the ranking are modelled figures from a commercial search-data provider, not measured traffic. We publish exactly how they are produced and the four things they cannot tell you: where our search numbers come from.

How is an answer captured?

This is the part most tools are vague about, so here is the method in full.

Each question is asked in a fresh temporary chat, with no account history, no memory and no personalisation carried between questions. That is the bias control: an assistant that remembers you asked about the same business four questions ago is not answering the way a stranger’s assistant would. The answer is reconstructed from the live response stream as it arrives, and then re-parsed authoritatively on our side.

From that captured answer, extraction is deterministic — rules, not a second AI asked to summarise the first. No model is ever asked to self-report whether it recommended you, because a model’s account of its own answer is not evidence.

One distinction is worth understanding, because it is where most competitive tools quietly overstate. When Radar links a cited source to a business, it keeps two separate buckets: referenced, where the cited domain is that business’s own site and the link is a fact; and inferred, where a citation is attributed to the one business named in the sentence it anchors. Inferred edges are labelled as estimates everywhere they appear, and where the attribution is ambiguous the system abstains rather than guessing. An estimate that does not look like an estimate is a lie with a number in it.

The same honesty applies to incomplete runs. Every run carries its status, the engine and model used, the pack version, and how many questions were expected against how many came back. Each individual question records whether it succeeded, returned an answer with no citation, returned an answer where your brand was absent, or failed outright — because “the AI answered and did not mention you” and “we did not get a usable answer” are completely different findings and must never be averaged together. Your visibility score is computed over answered questions only, and the dashboard always shows the denominator.

What you see week to week

AI Traffic is included from Boost upward; the rest of Track is in every tier.

Which AI engines are checked

The second number in each row is your custom question allowance. Boost and above also include the done-for-you fixes — Google Business Profile and Web Readiness — which is the point at which cite0 stops handing you a list and starts implementing it.

Radar compared, feature by feature

Below is cite0 set against each of the three platforms most often shortlisted alongside us, one capability at a time. Everything in these tables was read from each vendor’s own pricing, product and documentation pages on 25 August 2026; the sources are listed at the foot of the page.

cite0 vs Profound

Profound is a serious enterprise instrument, and the two products answer different questions. Profound tells a national brand how it appears across many markets. cite0 tells one business how it appears in the market it actually sells in — and then changes the profile, the site and the copy behind that answer inside the same subscription. At the same $99 entry, you track twice the questions.

cite0 vs Otterly.ai

Otterly is the widest engine coverage per dollar in the category, and if you are a national brand watching four engines at once it is a strong buy. The difference for a local business is where the question is asked from. Otterly targets a country; cite0 targets the neighbourhood your customers actually live in, which is the level at which “best med spa near me” is decided.

cite0 vs Peec AI

Peec has built genuinely interesting things — product-level shopping tracking and brand-perception clustering among them — aimed at brands with catalogues and multi-country footprints. For a business with one address and one phone number, the axis that matters is depth in a single market, and that is the axis cite0 is built on.

What you get here and nowhere else

Four things separate cite0 from every platform in those tables, and each is checkable rather than a claim.

• The whole loop in one subscription. Measure the answer, fix the profile, fix the site, fix the copy, then measure again. Every monitoring platform above stops at the dashboard; every local SEO platform stops before the AI answer. cite0 is the only place both halves meet, which matters because a fix you cannot measure is indistinguishable from no fix at all.
• Neighbourhood-level questions, pre-run. The other platforms target a country. We run curated question packs per category and per market before a client signs up, so your first report already knows who AI names in your neighbourhood. Nobody else holds a pre-run local baseline.
• Published weights, everywhere we score. Ten profile pillars, nine content checks, four severity bands on the website audit — every number on this site is traceable to a weight you can read and argue with. Across every competitor we examined, not one publishes the weights behind its score.
• Fixes are never the paywall. Website findings render into paste-ready snippets deterministically — no model call, no credit, all of them at once. You pay us to implement and to keep measuring, never for permission to see the answer.

What we commit to

Two things worth saying plainly, because the category is noisy about both.

We sell measurement and movement, not guarantees. There is no submission form and no paid placement in an AI answer, so nobody can promise you one. What we commit to is measurable movement on the signals AI demonstrably reads — your profile, your reviews, your listings and your site — with the before and the after both on the record.

Every number carries its provenance. Visibility is computed over answered questions with the denominator always shown. Citation-to-business links are stored as exact or estimated and labelled as such. The demand figures that rank your questions are modelled estimates from a commercial provider, and we publish exactly how they are produced. You can audit every figure we show you.

How often does Radar run the questions?

Weekly. Each cycle re-asks your full tracked set in fresh temporary chats and stores the result as its own run, with a fingerprint that flags a re-collection so nothing is double-counted. Weekly is the shortest cadence at which the movement is signal rather than noise; running the same questions daily produces a chart that looks busy and tells you less.

Can you guarantee ChatGPT will recommend my business?

No, and you should be wary of anyone who says otherwise. There is no ranking API, no paid placement and no submission process for AI answers. What is real is that AI answers are assembled from sources — your Google profile, your reviews, listings, local coverage, and whether a crawler can read your website at all — and those sources can be measured and improved. cite0 commits to measurable movement on those signals and to showing you the before and after. It does not commit to an outcome that is not in anyone’s control.

How is this different from a keyword rank tracker?

A rank tracker answers “where does my page sit in a list of ten blue links for this keyword”. Radar answers “when a person asks a question in their own words, is my business named in the reply at all”. The mechanics differ completely: there is no position one to occupy, the phrasing is conversational rather than keyword-shaped, the answer is assembled from cited sources rather than ranked pages, and the same question can produce different names on consecutive runs. That last property is why Radar is built around repeated sampling and honest denominators instead of a single daily position number.