Engines

An engine is one AI product a buyer can ask. Heralded normally asks six on every plan. Four answer every day:

• ChatGPT
• Gemini
• Google AI Overviews
• Google AI Mode

Perplexity and Microsoft Copilot answer twice a week: in the weekly report and in the daily update three days later. A brand’s first report doesn’t ask them, so it never waits on their slower answers; they answer from their first twice-weekly read.

​

Claude

Claude is an add-on for the whole workspace, and once you have it, every segment is read on Claude. It uses no extra prompt slots. Claude answers once a week, in the weekly report. Its price is on heralded.ai/pricing.

​

Paused engines

Heralded can temporarily pause an engine while its provider is unavailable. New runs skip that engine. Your Snapshot report and Home name it as paused for that run. Its unanswered prompts stay unmeasured and do not count against your score. Runs already in progress continue as planned. Resuming an engine affects new runs only; earlier reports keep their paused state.

​

The Engines page

Open Engines from More in the rail or on your phone. The page compares every engine your brand is asked on. Its sentence names the engines you score best on and worst on, listing every engine that ties, and says missing where your score is 0. Once there is an earlier weekly report to compare with, it adds any real movement, or None moved. With one engine, there is no sentence. Filter by segment, period, type, intent or topic. Each row shows the engine’s Score and band-coloured bar, Trend, Mentioned, Recommended, Understood, Praised and Answers. An engine asked weekly says Asked once a week under its name. A figure under the minimum shows ”–”; hover it for Not enough answers (). Understood and Praised read only the answers that name you, so an engine that never names you shows Not named; hover it for “Not named, so nothing to judge”. Across measurement pages, filter choices with a count of 0 appear greyed out after other choices and cannot be selected. Choices that clear a filter remain available. Score starts highest first. Select Engine, Score, Trend or a figure header to sort; select it again to reverse the order. Engine starts A to Z, Trend starts with the biggest rise, and figures start highest first. On your phone the table keeps Engine, Score and Trend.

​

Inside an engine

Open a row to see that engine’s Heralded Score, the answers and prompts behind it, the four figure tiles, weekly score bars and competitors. Open a tile to read the figure with this engine fixed in its scope. The competitors list pages through the server results. While a page loads, its previous rows stay visible and cannot be opened. Open a competitor with the engine, period and filters kept in its pane’s scope, including the prompt when the engine pane is fixed to one. Strongest prompts and Weakest prompts list up to three each, with their scores and trends on this engine. Open a prompt for its pane fixed to the engine. All prompts on opens Prompts with the engine filter selected. Sources lists the five hosts this engine cites most in the period, with their types, relationships and citation counts. Open one for its pane fixed to this engine, or follow All sources cites to Sources with the engine filter selected. Show answers opens the answers behind the pane with its engine and filters selected. An engine card inside a prompt opens the same pane fixed to that prompt. Its score, figures, competitors, sources and answers follow both subjects. The strongest and weakest prompt lists and the link to all prompts are omitted.

​

Google AI Overviews

Google doesn’t show an AI Overview for every search. When it doesn’t, Heralded records that the engine returned nothing, which is different from an overview that leaves you out. The answer is not measured and never counts against you; see which answers count. Expect AI Overviews to show Not measured more often than the other engines.

​

Why your own results differ

Ask ChatGPT yourself and you may get a different answer from the one Heralded shows. Replies vary from one ask to the next, and your own account can carry your history and location. Heralded asks the same prompts every day and reports the share of answers that name you, so any one reply you see is a single sample of what it measures.

​

How engines are weighted

Within each prompt, every engine that answered carries equal weight, however often it answers. Mentioned and Recommended explains the averaging. Adding or removing an engine changes what a figure is made of, so change is withheld until the window no longer spans it.