Buyers ask AIs to compare products more than to discover them. Comparison asks the AIs that question for your brand and shows how often they pick it. Its section in the project menu has four pages: Overview, Products, Attributes and Personas.
Products, attributes and personas#
A product is something your brand sells, such as "SEO consulting". Each product has its own competitors, taken from the project’s competitors or added by hand (up to 10), and its own attributes (1 to 8), chosen from the project’s attribute library, where attributes sit in groups you reuse across products. You may import Semantic Branding’s attributes; they are never added on their own. Personas are the kinds of buyer you sell to: three are suggested from your site, and you can edit them and add up to 10.
Generate products and attributes, on the Products page, reads your site’s home page and suggests products, the attributes of each (shared between products when it makes sense) and personas. Nothing is saved until you pick what to keep. The New project wizard offers the same step, Comparison, which you can skip.
The question#
Each period, every AI is asked one question per product and persona: "Which {category} is best for {persona}? Options: {options}." A product is named as you sell it (a product line, or a service by what it is) and carries its category, what a buyer asks the AI about. For each attribute it ranks the options and gives each a score from 0 to 100, with one checkable fact about that option. The scale is the one Semantic Branding uses: 60 or more only with a fact, 80 or more only when the option stands out, 90 or more for one option per attribute. The options are your brand and that product’s competitors, in a new random order in every question, and none is marked as yours.
Lastfind copies each score and its fact out of every answer, then checks it: a score of 60 or more with no fact behind it counts as 59. No answer is discarded: the places follow from the scores. When an AI gives no answer, fails, or answers without any score, the question is asked again, so every question ends with scores; a missing score is a dash, never a zero or a loss.
What the screens show#
Every figure comes from the attribute scores, so they never contradict each other. In each answer the overall score of an option is the mean of its attribute scores; the places and the winner, the single highest score, follow from it. The screens show the mean score, the win rate (the answers your brand wins, out of those where it was an option) and the average position, overall, on each attribute, for each product and each persona, beside every competitor. Each product has its own page: its scores against each competitor, their evolution in the period, the fact behind each AI’s score and the answers themselves. Overview filters by product and persona; the header’s AI and period apply everywhere. A figure without answers shows a dash, never zero.
AIs, frequency and limits#
ChatGPT, Gemini and Google AI Mode are asked, among the AIs your project monitors. Google AI Overviews answers a search and cannot take instructions, so it shows as not asked, with that reason; an AI not served in your project’s country is left out, and the screen says why. The questions follow your project’s collection: once a day, or once a week on a weekly plan (Settings > Monitoring). A product or persona you add is asked in the same period. Each question counts against the project’s monthly budget.