Conductor vs Scrunch for Measuring AI Search Visibility

How do they compare?

From each vendor's pricing and feature pages:

What does Conductor measure?

Conductor's AI Search Performance tracks brand mentions and website citations, share of voice against competitors, and sentiment tied to the sources cited. You can view results by topic or drill into prompts, and configure tracking by brand, region and domain. It also reports how AI bots crawl your site. Conductor relaunched the module in April 2026, calling it a system of record for AEO, and followed it with AgentStack, which adds APIs, an MCP server, apps inside ChatGPT, Claude and Copilot, and ready-made agents that turn findings into drafts.

Conductor sells plans by usage rather than seats. Essentials, which can be tried free, lists content drafts, keyword tracking and website monitoring but no AI Search credits. Growth includes 2,500 AI Search credits a year. The pricing page does not say what one credit buys, so ask how many prompts, engines and runs that covers before comparing costs. How it compares with another enterprise platform is covered in BrightEdge AI Catalyst vs Conductor.

What does Scrunch measure?

Scrunch's Core plan covers 125 prompts and 5,000 responses a month on ChatGPT, Perplexity, AI Overviews and Copilot. That is enough for each prompt to run on all four engines about ten times a month. It reports mentions, citations, competitor share and sources, with one country, persona and language on Core. It also shows AI agent traffic to your site and, for retailers, performance at product level. Enterprise adds five more engines, AXP, an API and Looker Studio reporting.

Scrunch also publishes research from its data. With Stacker, it found an average citation half-life of about 4.5 weeks across 3.5 million citation events, meaning half of citations to a page fade within about a month. That matters for how often you measure. More detail on the product is in Scrunch vs Peec.

How do their prompt approaches differ?

This is the biggest methodological difference. Conductor generates synthetic prompts from more than ten years of keyword data, shaped by buyer personas and intent. It argues that search volume figures for AI prompts are unreliable, because they come from small panels rather than real searches, and it does not show them. Scrunch starts from the prompts and personas you define, with fan-out to related questions on higher plans.

Neither approach is wrong. Keyword-grounded prompts tie AI tracking to known demand but may miss how people phrase questions to assistants. User-defined prompts reflect your buyers if you know them well, and your blind spots if you do not. Whichever you use, review the prompt list with sales and support teams, who hear real questions every day.

Can either measure whether AI selects you?

Mentions and citations show that an assistant knows you. Selection is the next step: the assistant recommends you as the answer, or an agent picks your product to buy or book. Neither Conductor nor Scrunch publishes a method for measuring selection as a separate metric, as far as their public pages show. Conductor's market share and Scrunch's position data come closest.

Measuring selection is harder than it looks. Deciding whether an answer "recommends" a brand is a judgement, and different reviewers, or different AI classifiers, often disagree. A workable approach uses strict, countable rules:

1. Use decision prompts only, such as "which should I buy for..." or "book me a...", not general questions.
2. Count first position, where your brand is named first or alone as the answer.
3. Count explicit recommendations only when the answer uses clear wording, such as "the best choice is" or "I recommend".
4. Run agent tasks by hand in assistants that can shop or book, and record what they choose.
5. Report ranges, because answers vary between runs.

The link between these rates and buyer behaviour is covered in how mention rate and top-three rate relate to buyer behaviour, and agent-led buying in how to prepare for agentic commerce.

How often should you measure?

Because answers vary from run to run, one day's result tells you little. Judge movement on averages across a week or a month. Repeated runs of the same prompt are correlated, so extra runs narrow the margin less than you might expect; to detect smaller changes, add more distinct prompts rather than more runs of the same ones. Scrunch's half-life finding suggests reviewing citations at least monthly, since a page can lose half its citations within about a month. For board reporting, monthly or quarterly figures are enough, as long as the prompt set, engines and competitor list stay the same between periods. Changing any of them breaks the comparison, so record every change alongside the numbers.

Which should you choose?

• Conductor if you want AI visibility inside a wider SEO and content platform, with prompts tied to keyword demand and agents that act on findings.
• Scrunch if AI search is the focus, you want a published entry price, and you may want to serve AI-ready pages to agents through AXP.

Both suit enterprises; Scrunch's Core plan also fits mid-sized teams. Test either on the same 20 decision prompts, and check a sample of answers by hand.

What might this look like in practice?

Consider an illustrative project management software company, Jaskmere. Scrunch showed it mentioned in 60% of tracked answers, but a manual review of 20 decision prompts found it named first in only three. The team added a selection column to its monthly report, using the strict rules above, and focused work on comparison pages and third-party reviews for those prompts. Six months later, first-place selections had risen from three to seven of the 20 prompts, while overall mentions barely moved.

What do people ask about Conductor and Scrunch?

Which tracks more AI engines?

They are level. Conductor's documentation lists nine engines across more than 160 countries, and Scrunch Enterprise lists nine, including Meta AI; Scrunch Core covers four.

Does Conductor show AI prompt volumes?

No. It says reliable volume data for AI prompts is not available and generates prompts from keyword data instead.

What is Scrunch AXP?

A layer that serves simplified, AI-ready versions of your pages to AI agents at the same URLs.

Can either tell me if ChatGPT would buy my product?

Not as a published metric. Test agent tasks by hand and track first-position recommendations.