Prompts & Rankings

Overview

Prompts are the foundation of everything in AEO Optima. A prompt is a question or query that gets sent to AI answer engines — the same kinds of questions your potential customers are asking ChatGPT, Claude, Gemini, Perplexity, and other AI tools every day.

By defining the right set of prompts, you control exactly what you're monitoring. Every snapshot, every visibility score, and every trend line traces back to the prompts you've configured.

Why Prompts Matter

When someone asks an AI model "What is the best project management tool?", the AI generates a response that may or may not mention your brand. AEO Optima lets you define that exact question as a prompt, send it to multiple AI models on a regular schedule, and track whether your brand appears — and where it ranks — in the responses.

The quality and coverage of your prompts directly determines the quality of your visibility data.

Creating Effective Prompts

Different prompt types reveal different aspects of your AI visibility. Use the table below as a guide when building your prompt library.

Prompt Segmentation (Branded vs Non-Branded vs Competitor)

Every prompt is automatically classified into one of three segments based on whether it mentions your brand name, competitor names, or neither:

Why Segmentation Matters

Without segmentation, branded prompts inflate your visibility scores. A query like "What is [Your Brand]?" will almost always mention your brand — that's not organic discoverability, that's expected. By filtering to Non-Branded prompts only, you see your true organic AI visibility: how often AI recommends you when users aren't specifically asking about you.

How Auto-Detection Works

When you create or import a prompt, AEO Optima automatically:

1. Checks if the prompt text contains your brand name (fuzzy matching handles abbreviations and variants), an alternative name you saved for your brand, or your brand's short name (for example "Acme" for "Acme Foundation"). A short name that appears only inside a competitor's name does not count, and a form you marked as not your brand never counts.
2. Checks if any competitor names from your competitor configurations, or the alternative names you saved for them, appear in the text.
3. Detects "vs" or "versus" patterns that indicate a comparison query.

The detected segment is shown as a badge on the prompt. You can always override the auto-detection by manually selecting a different segment.

Saved alternative names and the short name are used from 22 October 2026. Prompts created before then keep the segment they already had.

Using the Segment Toggle

Every analytics page includes a Segment Toggle at the top: All | Branded | Non-Branded | Competitor. Switching segments filters all charts, KPIs, and data on the page to show only data from prompts in that segment. The toggle is URL-synced, so filtered views are bookmarkable and shareable.

Segment in Reports and APIs

• Reports: Choose a segment when generating a report to focus on branded-only or non-branded-only metrics.
• API: All analytics endpoints accept a ?segment=all|branded|non-branded|competitor parameter.
• CSV: The prompt_segment column is included in CSV exports and recognized during imports.

Setting Business Priority

Every prompt can be assigned a Business Priority score from 1 to 10. This score helps you focus your attention on the prompts that matter most to your business outcomes.

Priority scores are used to weight visibility calculations and to sort prompts in reports, so your most important queries always surface first.

Topics

Every prompt can have a Topic — a short label used to organize and group related prompts. For example, prompts like "What is the best CRM?" and "Top CRM tools for enterprise" might both have the topic "CRM Software".

• Topics can be set manually when creating or editing a prompt.
• If left blank, a topic is auto-generated from the prompt text by extracting the most meaningful words.
• The prompts table shows a color-coded topic column for quick visual grouping.
• A Topic filter dropdown lets you view only prompts belonging to a specific topic.

Managing Your Prompt Library

Every change to the tracked question set is recorded — when it widens (a question added or imported), narrows (a question removed), or a question is refiled from one type to another — so a rate that moved because you changed what you measure can be told apart from a rate that moved because your visibility changed.

Adding a Prompt

1. Navigate to Prompts from the sidebar.
2. Click Add Prompt.
3. Enter the prompt text exactly as a user would type it into an AI chatbot.
4. Optionally enter a Topic for grouping (auto-generated if left blank).
5. Set the Business Priority (1-10).
6. Optionally set the Prompt Type, Search Volume, and Page URLs.
7. Optionally tick What this question is about: the contexts on your context map this question is about. The findings read the kind of work from these ticks and never guess from the wording. A question opened from a context gap's Add a question arrives with that context ticked; untick it if it does not fit. The Add question dialog on the Query list offers the same ticks.
8. Click Create Prompt.

The standard questions about your brand

At the top of Query list & types, owners, admins and members see Add the standard questions about your brand. Opened, it lists questions buyers ask about a brand, written from your brand's name and its first competitor: whether it is a good brand to buy, its reviews, how it compares with a named competitor, complaints, and more. Keep, edit or untick each one. A question with no competitor to name opens blank, so you can fill it in or leave it unticked. A question you already track is shown as tracked, with a link to it. When one of your own questions already says the same thing, it is shown as In your questions (counted) and is not added a second time.

Include the identity check adds three fixed questions: what your brand is, who owns it, and what it makes. They check that answer engines recognise your brand correctly (gate 1, recognised correctly). The three are added together or not at all, and their wording cannot be edited, so every cycle asks the engines the same question. The card names the brand they will ask about, and where to change that name first (on the Context map, Define contexts, first step Who we are). If one of your own questions already asks one of the three, it becomes part of the identity check: no second question is added, its answers stay with it, and from then on it is not counted against your question limit. Gate 1 can only be read once your owner (or Independent: no parent company) and what your brand makes are given. The card asks for whichever is missing, and the questions are asked every cycle either way.

When standard questions are switched off, the card says how many and how many more answers each run would add, from your project's own engines and locations. Owners and admins can press Switch them all on. Questions that don't fit your plan's limit stay off, and the card names them. The three identity questions go on together or not at all.

These questions are not counted against your question limit. They are asked every cycle like any question, so they use answers. The three identity checks are a test of whether engines recognise you, not a measurement of how they treat you: their answers are kept and you can read them on the Answers page, but they are in no rate, count or tone anywhere else, and never in a scorecard's frozen set. The answers download and the raw responses export mark each of those answers in an Instrument column, so a file can be reconciled against your report. The ten other questions count like any question. Any that your plan's answers don't stretch to are added switched off. They are not offered on the Free plan, whose daily answer limit can't complete one cycle of them. On the list they carry an Identity check or Standard brand question badge.

Importing from CSV

You can bulk-import prompts from a CSV file. This is the fastest way to add a large number of prompts.

1. Click Import CSV from the prompts header.
2. Download the template or drag-and-drop your own CSV file.
3. CSV format: query, topic, tags — only the query column is required.
4. Review the preview — valid, duplicate, and error rows are highlighted.
5. Click Import to create all valid prompts.

Every imported prompt is auto-classified with intent, volume, SDS tier, journey stage, freshness, risk level, and topic. No manual classification needed.

A row that repeats one of the standard questions about your brand is never imported: it is already tracked, outside your question limit, and a second copy would be counted and asked twice. The preview marks it Standard question, and the result says how many were skipped. Adding a question by hand, or through an AI assistant, follows the same rule, and so does editing one: new words that repeat a standard question are refused, while a standard question saved in its own words is kept.

Exporting to CSV

Click the Export button (download icon) to download all prompts as a CSV file with full metadata including query, topic, tags, type, intent, journey stage, volume, SDS tier, freshness, risk, and active status.

Editing and Archiving

• Click any prompt text in the table to inline-edit it directly.
• Use the row action menu to open the full edit page, take a snapshot, or delete.
• Prompts can be deactivated with the Active toggle rather than deleted, preserving all historical snapshot data.

Organizing by Topic

Topics help you group related prompts for easier filtering. Common topic schemes include:

• By product line: CRM Software, Email Marketing, Analytics Platform
• By competitor: HubSpot Comparison, Salesforce Alternative
• By use case: Lead Generation, Customer Retention, Onboarding

Best Practices

Tip: Start with the prompts that cover your most important use cases (how many your plan tracks depends on the engines switched on; the Query list & types page says when a project no longer fits), or use Import CSV to bulk-add prompts and let auto-classification handle the rest. You can also use Discover Prompts to generate prompts with AI assistance.

Write Prompts Like a Real User

AI models respond differently to conversational queries than to keyword-style searches. Write your prompts the way a real person would ask a question:

• Good: "What is the best email marketing platform for small businesses?"
• Less effective: "best email marketing platform small business"

Cover Multiple Angles

For your most important topics, create prompts from different angles:

• Direct brand query: "What is [Your Brand]?"
• Comparative: "How does [Your Brand] compare to [Competitor]?"
• Problem-oriented: "I need a tool that does [thing your product does]. What should I use?"

Review and Rotate

AI model responses evolve over time as models are retrained. Review your prompt library monthly:

• Remove prompts that are no longer relevant to your business.
• Add prompts that reflect new products, features, or competitive dynamics.
• Adjust priorities based on what's driving the most business value.

Monitor the Rankings

When your brand appears in a numbered list within an AI response, AEO Optima records your rank position. Track this over time — moving from position 5 to position 2 is a meaningful improvement, even if your overall visibility percentage stays the same.