Where Should Your AI Search Prompt Set Come From? A Practical Comparison of SAG, Profound, and Semrush

A visibility report can be precise and still answer the wrong question. Before comparing brands in AI answers, teams need to decide where the prompts come from: the questions their customers actually ask, patterns discovered across conversations, or a prioritized list the team wants to monitor. Those approaches overlap, but they are not interchangeable.

This comparison focuses on prompt-set provenance and decision rights, not a ranking of accuracy or outcomes. Product descriptions below were checked against each provider’s official public pages on October 19, 2026. Features, coverage, and limits can change; confirm the linked documentation before purchase.

The short answer

• SAG starts with customer questions selected with the organization and compares the company with up to five competitors under those question conditions. It is a question-led diagnostic and improvement workflow.
• Profound Prompt Volumes describes discovering prompts from real AI conversations, then organizing and prioritizing them by topic, intent, demographics, and volume for tracking.
• Semrush AI Visibility Toolkit combines prompt/topic research with the ability to select specific high-value prompts for day-by-day tracking. Its documentation distinguishes research and tracking as separate tasks.

The first practical decision is whether you need to define the questions, discover candidate prompts, or monitor a chosen set. Many teams will need more than one of these steps.

Comparison: who defines the questions, and why?

Important distinction: “discovery” does not mean the same thing as “measurement.” A discovered prompt is a candidate. A customer-defined question is a scope choice. A tracked prompt is a monitoring commitment. Record which category each item belongs to before interpreting a trend.

What SAG is comparing, and what it is not

SAG’s public workflow describes setting a site, customer questions, and up to five competitors, then comparing answers question by question. Its measures are deliberately separated: SEO checks search exposure; AEO checks whether the brand is mentioned in answers; GEO checks whether an official source is cited. The workflow proposes improvements and supports rechecking the same questions and conditions after changes.

That makes SAG’s primary unit of work a customer question under an agreed comparison scope, not a claim that one universal prompt list represents every market. The Cream Union describes its digital experience work as dating to 2004. SAG offers SaaS and on-premises choices, while its live service page says site modification work is separate. Teams should agree in writing who approves and implements changes, what collection is included, and which costs or integrations sit outside the selected service.

A practical prompt-set design that combines the approaches

A team does not need to choose one source forever. It can create a governed set with three labeled lanes:

1. Customer-validated questions: Start with questions heard in sales calls, support tickets, search queries, and customer research. Remove personal information and group duplicates. Ask customers or subject-matter experts to confirm that each question reflects a real decision.
2. Discovered candidates: Use conversation-derived or topic research to find unfamiliar wording and adjacent needs. Treat these as hypotheses until someone validates the audience, geography, intent, and relevance.
3. Tracked priorities: Select a stable subset tied to a decision, such as a launch, an important service category, or a recurring customer concern. Keep the wording and conditions versioned. If a prompt changes, record the change instead of presenting the new series as directly comparable to the old one.

For each prompt, store: source, exact wording, language, market, intent, owner, competitor scope, platforms checked, first-seen date, validation status, and reason for tracking. Keep one stable core set for trend comparisons and a separate discovery queue for new candidates. This avoids silently mixing exploration with performance reporting.

Buyer checklist: questions to ask in a demo or contract

• Can we see the exact prompt wording and its source, or only a topic label or aggregate score?
• Can our team add, edit, approve, and retire prompts? What is the review process?
• Are customer-provided questions and system-discovered prompts distinguishable in exports and reports?
• Which languages, countries, answer engines, and locations are actually included for our plan?
• What does “daily” or “weekly” mean: refresh time, rolling window, or a new collection run?
• Can we compare results only when question, platform, locale, and collection conditions match?
• How are missing answers, uncited answers, and zero mentions represented? Do not assume a missing observation is zero visibility.
• Who turns findings into approved page changes, and is implementation included or a separate service?
• For on-premises deployments, which server, GPU, API, integration, maintenance, and support costs are excluded from the quoted scope?

Choosing a workflow

Choose a customer-question-led diagnostic when the decision begins with a known audience and a specific competitive question. Choose conversation-based discovery when your team needs to expand or challenge its assumptions about how people phrase needs. Choose prompt tracking when you already have a justified, stable shortlist and need to monitor it over time. These are workflow distinctions, not a universal performance ranking.

A useful pilot is small: agree on a core set, label where each item came from, define competitor and market scope, save the exact conditions, then review whether each result led to an actionable content or technical change. Recheck the unchanged core set after implementation; evaluate newly discovered prompts separately. This gives decision-makers a traceable path from question selection to action without treating a tool’s prompt volume or visibility score as a guaranteed business outcome.

Official sources checked October 19, 2026

• SAG service workflow and scope: SAG — customer questions, competitor comparison, improvement, and recheck · Service onboarding FAQ · Measurement FAQ · SaaS and on-premises FAQ
• The Cream Union: Company homepage — digital experience since 2004
• Profound: Prompt Volumes
• Semrush: AI Visibility Toolkit — official knowledge base

This is a comparison of publicly described workflows, not a controlled benchmark. Public documentation does not establish that the products use identical prompts, models, locations, or collection conditions; no result or ranking is implied.