The page
The headline describes how much of what AI says about your brand is right. The Understood and Praised cards and the lists below them follow the period, engine, type, intent, topic, market and audience you select. When a measured figure has no earlier comparison, its card says First report. The Facts table lists your claims and facts, most contradictions first. Each statement lists its parts with a quoted answer, its non-zero verdict counts and the change in contradictions. A part that no answer mentions is marked Gap and keeps its never-say count. Other parts show Not enough answers yet until three answers address them. A fact keeps its own row. Quotes are plain sentences from prose, never a table row or a heading. With no suitable quote, the counts still show. A rise in contradictions is red and a fall is green. Open a row to see Agrees, Contradicts, Partly and Doesn’t say, the pages that contradict it, the lines answers said in their own words, and work on it. Open as a page adds each engine’s contradictions and the questions behind them. The closing link opens the answers that contradict that wording under the pane’s filters. Under Where it goes wrong, Questions lists questions by their share of negative answers naming you. Sources lists cited pages that speak against you, most cited first, with passages served as plain sentences. Best quotes opens favourable answers word for word, leaving out headings, table rows and fragments with fewer than five words after a leading label. Each quote ends with sentence punctuation. Page addresses do not count as words. A competitor’s Positioning cell opens its pane. Each figure card counts the answers it reads, and the pages too when it weighs them. The figure cards and negative question shares show Not enough answers with the answer count when the sample is too small.Understood
Understood measures how much of what the answers say about your brand is right. It reads four statements, the ones you set under Identity in your brand settings:- Category, what kind of product you offer
- Audience, who you serve
- Offer, what you do for them
- Difference, what sets you apart
Brand facts
Your brand can hold up to 20 facts, each a short statement of up to 300 characters, with who added it and when. Every update checks each fact against each answer to your tracked prompts that names your brand. Every weekly report also asks each fact’s own question, such as “When was MOCAPP founded?”. These questions are not tracked prompts. They use none of your prompts, and their answers count toward that fact alone: never toward Mentioned, Recommended, Understood, Praised, your sources or any other figure. A fact with no question, such as an opinion, is checked only in your tracked prompts’ answers. The verdict is agrees, partly, contradicts or doesn’t say. Where an answer supports or contradicts a fact, the reading keeps the passage that decides it. A fact’s row counts the answers to its own question and to your tracked prompts together, and the API reports them apart. From the first weekly report that asks a fact its own question, its trend starts a new baseline until the reports it compares are all ones that asked it, so the added answers never show as a change. A fact with a check is judged by it. The check is the comparison a right answer meets, with its units, scope and qualifiers. “More than 8,000 creators” agrees with “7,500+ influencers” but only partly agrees with “7,500+ vetted influencers”, because it says nothing about vetting. Cited pages and the passages that contradict a fact are judged by the same check. Without a check, a plain quantity such as “8,000 creators” means about that number. An answer agrees within 10%. An explicit comparison such as “at least 7,500 vetted influencers” keeps that comparison. Dates stay exact and ranges keep both ends. Agreement also requires the fact’s scope, time, units and qualifiers. Saving or changing a fact checks it against every existing fact, including claims inside longer statements. A duplicate is refused without adding a second copy. A contradiction leaves your facts unchanged until you choose which claim to keep. If that check is unavailable, nothing is saved. Studio also checks a replacement against the other facts. Choose Add a fact under Strategy › Facts. Perception and Brand settings link there too. Check previews the separate facts, their checks and their questions before you save. Untick any you do not want. An opinion shows Not measurable with a suggestion instead. When a fact contradicts another, choose Keep both, Replace or Keep the existing one, then Save choices. Existing facts keep your wording. A single fact gets its check and question. A longer statement keeps its whole-sentence measurement and offers a split, such as Split into 2 facts?. Review the suggested facts, edit or untick them, then confirm to replace the original in one save. Closing the dialog keeps the original. If your saved facts change before you confirm, choose Check and review the refreshed selection. Interpreting, Could not interpret, Not measurable and Brand paused distinguish facts awaiting interpretation, an unavailable provider, opinions and paused brands. You can also save a statement you confirmed on a piece in Review, as statements to confirm describes. A fact saved from a piece shows From and the piece’s title. In the fact panel, Edit wording opens an inline field with Save and Discard. Brand settings also lets you edit the wording or remove a fact. Remove fact asks you to confirm before removing it. Cancel keeps the fact. New wording is checked from the next update on. The writer uses your facts too, as its orders describe. Cited pages are also checked against your facts, using the passages around your brand’s mentions. A flagged page keeps its contradicting passage beside the fact’s wording. A flag is evidence to review and does not open an action by itself. Claim flags on your own cited pages keep the passage and wording from that update. A page check can finish after the report is delivered. Its verdict and passage appear when the check completes. Later edits or recrawls do not change that completed evidence. You can file a correction request for a third-party page or a content update for one of your own pages, for a claim or a fact. The action keeps the flagged passage and the wording it contradicts. Each page and claim or fact has one action, and its open work appears in the pane. Compare The page says with You say before choosing Work on this, which offers a correction request for a third-party page or a fix of your own page. Once the work is filed, the page shows its stage. The page title opens its page panel. Readers without permission see neither Work on this nor Not a contradiction. Not a contradiction dismisses that page’s flag for the claim permanently. It leaves the pane’s page list and page count. If the request fails, the page returns. A fact keeps its history across edits. Fixing a typo, or rewording it without changing what it claims, keeps every earlier reading counted and comparable. Changing a figure, a date, its scope or a qualifier starts a new version: its row counts only the answers read against that version, and its trend starts a new baseline with no change shown across the edit. Until the next update reads the new version, the row keeps the previous version’s counts, labelled Earlier answers. The current fact shows its next check date or Waiting for new answers. Earlier answer and page readings keep the wording and check they were read against. Removing a fact stops future checks and preserves earlier readings. Facts do not change Understood, which uses the four profile statements only.Praised
Praised is the share of answers naming you that frame your brand positively. A neutral or balanced description doesn’t count as praise.Labels you are moving away from
The labels you list under Labels you want to leave behind in your strategy get their own reading. A label such as “crowdfunding platform” is a description you no longer want AI to give. Every update asks, of each answer that names you, whether it describes your brand with the label or with words that mean the same, even inside a longer description. An answer that says you are not that, or uses the words only about another company or about the market, does not count as using it. The figure is the share of answers naming you that use the label. It counts only the answers read since you added the label, so a new label starts at your next update, and a label you reword starts a new reading. With fewer than three answers it shows no share. Like Understood and Praised, it is averaged within each question and engine and carries a 95% range. Its change gets the same paired test as the other figures, and a rise or a fall is named only where the test calls it real. Here a fall is the improvement. Every cited page that names you is read against each label too, your own pages included. The pages that describe you with a label are listed beside its figure, your own first and then the most cited. A page of yours on that list is the place to start. The labels do not change Understood, Praised or any other figure. The API and MCP return each label’s share, change, newest quote and pages.How the figures are averaged
Both use the same averaging method as Mentioned and Recommended. Understood first averages the answer scores within each question and engine cell. Praised first takes the share of positive answers in each cell. Each engine with a measured answer then weighs its usage share within its question, rescaled over the measured engines, and each measured question weighs equally in the figure. Understood and Praised can be lower or higher than every engine’s own figure because questions weigh equally. Only answers that name you and carry a reading count toward either figure. A cell with no reading drops out. In a multilingual market, completeness follows the answers returned in each language. An answer that leaves your brand out needs no Understood or Praised rating to complete that language. The answer count states how many readings the figure rests on; a question or engine with more answers gets no extra weight.How they differ from the other figures
- They only read answers that name you. An answer that leaves you out has nothing to describe. The app shows how many answers each figure was read from. See When a figure is empty for what appears when none named you.
- They include questions that name your brand. “Is Acme a good CRM?” says a lot about how an engine sees Acme, even though it can’t count toward Mentioned.
- They read the whole week. Like the other figures, they count the answers the weekly report pools, daily updates included.
- They carry a 95% range. They get the same paired test as the other figures, and Home shows a change when it is real.