Product journal

Tadow is now available in American English: screens, emails and help. Everyone picks their language in Settings, without changing it for their teammates.

What Tadow writes for your brand (Opportunities actions, generated texts, emails) follows the workspace’s “Content language”, which an owner or admin can change: it switches to English at the next scan.

Until now, an action came from a lost query. Everything else costing you visibility lived on other screens, or nowhere.

When AI reads Amazon or Sephora before answering, you know. When ChatGPT or Gemini never cites you, you get that engine’s own lever. When a rival buys your queries, you see what its ad promises. When AI speaks badly of you on a subject, you read its sentence.

Then there are the listings it cannot read, the ones with no structured data, and the llms.txt file your site does not have: we generate it, you publish it. Each of these produces an action, with its proof, the decision to take and three steps at most.

An action marked done is re-measured by the scans that follow. If nothing moved after two scans and two weeks, we say so rather than leave you waiting.

When AI recommends your brand less on a query, it is not always an organic loss. Sometimes someone bought the slot.

Before, since

For every advertiser, we compare your presence on the queries they buy, before their first ad and since. The worst hit query is named, with a link to open it.

What we refuse to say

Only queries tracked on both sides enter the calculation, and it takes at least eight measurements across two scans per side. Below that the screen gives the reason for its silence instead of a number: on a single query presence is 0 or 100 %, so a gap there would be chance.

An ad that is too recent reads « not measured since », never « no effect ». And when one advertiser’s queries move in opposite directions, we do not average them into a sentence none of them lives.

A coincidence, not a cause

The ad is captured on free ChatGPT, where it shows; your presence is measured through the API, where it does not. Those are two renderings of the same engine, and what we show is a coincidence in the calendar, not a demonstrated cause: other things may have moved at the same time.

Between your tracked query and the answer on screen, ChatGPT and Gemini run searches of their own. We were already capturing them; they now have a screen, in Overview, next to Citations and AI visits.

Two moments, two different jobs

The engine first names the brands it already has in mind, then opens the site of the ones it shortlisted. Those are very different things: missing from the first list, you are never even searched for, and no listing fix will change that. Read and then dropped, your listing failed the check, and that one is entirely yours to fix.

The screen sorts your answers into those three cases and gives you, whenever the engine read your site, the exact words it was looking for there.

The real words of your market

The ranking shows what the engine types, not what we guessed when writing the query. Below it, the sites it opens to verify, and your rank among them.

Perplexity and both Google surfaces do not expose their searches: their answers are still counted, and the screen says which ones it cannot read.

When AI recommends one brand over another, something decided it. Opportunities starts there: every query you lose produces a recommendation, and there are three kinds.

Technical

An AI crawler your robots.txt file keeps out of your store. While it is blocked, none of your listings can be recommended, however good they are. We give you the block to paste and where to paste it.

The criterion that decides it

AI compares brands on one precise detail, and the brand that states it wins the query. We read what it was looking for in the answer, check what your listing says about it, and name the detail that is missing. When it is a certification, we name the bodies to contact as well.

The pages AI goes to read

On some queries it never considered you at all: it went and read comparison articles and specialist sites that cite your rival and not you. We list those pages for you.

Every recommendation states the problem with the numbers, the cause, the decision to take and how to do it in three steps at most. It also says who carries it out, Tadow, you, or your agency, and you can hand us the ones that are not yours to do. Once it is marked done, the scans that follow re-measure its queries and the screen shows you before and after.

"1 mois offert", in the sidebar, opens your referral page: a link and a code that belong to you. It counts visits, brands signed up and months earned.

A brand coming through your link has no code to enter, the discount is already in its payment. If it only has the code, it pastes it at signup or just before choosing a plan, on our screens, not in Stripe’s field.

A month is earned on a paid subscription, not on a signup. We apply the credit ourselves and write to you once it is done. You never see the identity of the brands you introduce, only where each referral stands.