Twenty-one questions, and our answers

Georadar queries ChatGPT, Perplexity, Gemini and Claude, each one separately because their answers diverge: the same prompt may well cite you on Perplexity and ignore you on Gemini. Georadar keeps the engine-by-engine detail alongside the consolidated view.

Being cited means appearing in a list such as "several brands exist: X, Y, Z". Being recommended means being put forward, as in "I would suggest X". These two situations weigh very differently on a purchase decision, which is why Georadar counts them separately with a distinct mention rate and recommendation rate.

When an engine cites its sources, Georadar extracts them and aggregates them by domain, so you get a ranking of the sites feeding the answers on your prompts, with your own domains flagged. The editorial action plan starts there, since you then know which sites to aim for.

The reference cadence is weekly: one full reading per week, at the same moment, so the curves stay comparable. You can launch a reading on demand at any time, for instance just before and just after a publication.

Yes, and without declaring them in advance, which removes the classic blind spot of the competitor nobody was watching. Georadar extracts every brand cited in the answers, including those you were not watching. You promote a detected brand to tracked competitor in one click, and its history remains available.

The Radar Index is the Georadar composite score: a number from 0 to 100 summarising a brand's generative visibility. It aggregates six separately recorded axes, namely presence, recommendation, share of voice, primacy, stability and tone. Each axis carries a fixed and published weight, and the detail stays viewable axis by axis.

Because an isolated metric ranks real situations badly: a brand cited in 80 per cent of answers but always last, never advised and in lukewarm terms is not in the same commercial position as a brand cited in 50 per cent of answers, consistently first and actively recommended. The Radar Index fits those dimensions into one manageable figure without hiding the detail.

Yes, the six weights appear on the methodology page and are in no way confidential: presence 25, recommendation 25, share of voice 20, primacy 15, stability 10, tone 5. A score you cannot recalculate is of no use to you.

From three sources you combine freely: your Google Search Console queries imported by the connector and turned into conversational questions, automatic derivation from an analysis of your site, and your own input. Prompts are organised into categories, so they can be read by product world.

Yes, to both, and that is what makes the measurement plan relevant from the first week. Search Console feeds the measurement plan with your real queries rather than guessed prompts. GA4 closes the loop by matching assistant-referred sessions against your conversions, so generative visibility can be read against your other channels.

Through an in-house pixel placed on your pages, already deployed by default on articles published across our network. It recognises the declared assistant crawlers, namely GPTBot, PerplexityBot, ClaudeBot and Google-Extended, using a user-agent signature base we maintain, and it separates those visits from human traffic. A crawl today is a leading indicator of a citation tomorrow.

Yes, the measurement views export and the raw engine answers stay viewable prompt by prompt, so you keep a record of what was answered, when, and by which assistant.

A set of 70 thematic editorial sites covering 25 verticals, published and hosted by R-Advertising. We control the entire technical and editorial stack, so there is no middleman, no negotiation and no external sign-off delay. The detail of the domains and their audiences is presented in person, along with the sites selected for your own prompts.

Because assistants build their answers mainly from third-party sources, and because your site carries authority on who you are rather than on how the brands in your category rank. To weigh on an answer of the "which brand should I choose" kind, you need to appear in the comparative editorial content engines consult. The sources screen shows you, on your own prompts, which domains actually feed the answers.

The dedicated article is a full piece written around a specific prompt where your brand is missing, published on the most relevant site in the network. The insert is a short editorial block placed inside an existing high-traffic article, as an expert viewpoint, a data panel or a usable tip. The article works depth, while the insert works volume and speed to publication.

The delay depends on the engine queried. On assistants that browse the live web, such as Perplexity, AI Overviews and ChatGPT in search mode, the effect can appear as soon as the content is indexed and crawled, which the pixel shows you in real time. On answers produced without browsing, presence follows the models' training cycles. Georadar separates the two situations in the reading, so you know where the progress is happening.

These answers are not advertising slots and therefore cannot be bought, which holds across the whole market. Our commitment covers the mechanism, and it is specific: a reproducible weekly reading, a baseline frozen at the outset, the production and actual publication of the content, then page-by-page tracking of AI traffic. Your first campaign will tell you the rest, with figures to hand.

Creating the account takes a minute and asks for no card. Georadar then analyses your site to propose a first set of prompts that you approve or correct, and the first reading launches straight after. Results appear as soon as it finishes.

No, Georadar sits alongside your agency, since SEO works your position in the results pages while Georadar works your presence in the written answers. The two feed each other: prompts drawn from your Search Console come from SEO, and content published on the network is indexed like any editorial content.

R-Advertising, a French company that publishes and operates the editorial network exposed in the platform, with teams in Mougins and Aix-en-Provence. Measurement and activation therefore come from the same operator, which saves you having a tool on one side and a media middleman on the other.