Step 4: cited, recommended and well described
updated on 2026-11-18
For each question and each AI, Siggmund reads the answer the way your customer would. It does not just count citations: it looks at whether the AI recommends you, for whom, to the benefit of which competitor, and which sources it relies on.
In short
- Three situations to tell apart: absent, outranked, wrongly described. They do not call for the same actions.
- Four measures: your presence, the customer's reaction, the AI's sources, your reputation.
- The results are consolidated by a Bayesian statistical model, AI by AI.
The three situations
Not appearing, appearing behind a competitor and appearing with false information are three different problems. Mixing them into a single score means choosing the wrong action.
Your brand isn't cited
"Which sofa bed for a small living room?" The AI recommends three retailers. Yours isn't one of them.
A competitor is recommended
"Where can I buy a firm mattress?" The AI cites you, but third, and recommends a competitor first.
Your offer is misdescribed
"Does this retailer deliver to Belgium?" The AI says no. Your delivery page says otherwise.
Twelve questions, six AIs, one answer per cell
The sources the AIs used for these answers
Example: twelve questions from a furniture retailer, across the six AIs. Competitor names are made up.
What Siggmund measures
For each answer, Siggmund records four families of information.
Your presence
Citation rate, position in the answer, first-choice recommendation, AI by AI and need by need.
The customer's reaction
The recreated customer reads the answer and says whether they would go to you or to a competitor. Being cited isn't enough, you have to make them want to.
The sources
The pages, media, comparison sites, forums, resellers and marketplaces the AI relies on to answer, and which cite your competitors.
Your reputation
The qualities AI attributes to you, your customers' concerns (health, environment, ethics, quality, service) and repeated errors about your prices or services.
What do AIs say about your brand?
See what AI says about meThe sources: where the AI looks for information
When an AI answers, it relies on pages: media, comparison sites, forums, resellers, marketplaces. If those pages cite your competitors and not you, the AI will do the same.
Siggmund records these sources, question by question. That is what then lets you act where the AI looks for information, and not only on your site.
A statistical model, not an average
AI answers vary from one time to the next. A raw average over a few answers can suggest progress or a drop that does not exist.
Siggmund consolidates the results with a Bayesian statistical model designed by the team's PhDs in mathematics, AI by AI. You see where you are gaining, where you are losing, to the benefit of which competitor, and why.
Frequently asked questions
What is the difference between being cited and being recommended?
Being cited means appearing in the answer. Being recommended means being the choice the AI advises. A brand can be cited often and rarely recommended.
What does “wrongly described” mean?
The AI gives false or outdated information about you: a price, a delivery area, a service you no longer offer. Siggmund compares it with what your site says.
What do you see in the first baseline?
What AIs say about your brand, the competitors ahead of you and your first actions, in your Siggmund workspace.
Are the results separated by AI?
Yes. Each AI has its own results, because they do not answer the same way.