Why benchmark AI buying queries now

AI answer engines increasingly resolve commercial research without clicks, so brands must be discoverable, accurately represented, and cited inside those answers. Independent studies and Unusual’s field data show sharp declines in traditional search CTR when AI summaries appear and a growing reliance on third‑party sources in AI answers. Your share‑of‑voice (SOV) across AI systems is now a leading indicator for pipeline. See: Unusual on declining search traffic and zero‑click trends; Amsive’s analysis of CTR impacts and AI citation patterns; Beeby Clark Meyler and Bloomfire on structuring content for AI parsing. Search traffic is drying up, Amsive guide, BCM guide, Bloomfire overview.

What this benchmark covers

• Models: ChatGPT, Gemini, Perplexity, Claude. Unusual product overview and AI page.

• Query set: High‑intent buying prompts (e.g., “best [category] for [ICP]”, “[vendor] vs [competitor]”, “pricing”, “implementation time”, “security”, “[category] for [industry/size]”).

• Metrics: Model‑level SOV, mention rank/position, quality‑of‑mention (helpful/neutral/harmful), citation presence and source mix, correctness vs. your canonical facts, and change over time.

• Coverage sources: When AIs cite third parties (e.g., Wikipedia, Reddit, YouTube), quantify which ones drive your presence so you can target the right earned media. See external benchmarks on which sources AIs cite most. Amsive guide.

How Unusual measures AI SOV (methodology)

1) Define the buying‑intent question set with you, mapped to ICPs, use cases, objections, and compliance topics. 2) Query each model on a fixed cadence and capture the full raw answer, citations, and any links suggested for further reading. 3) Score: brand mentioned (Y/N), position salience, accuracy against your canonical facts, presence of owned citations (including your AI‑optimized subdomain) and high‑value third parties. 4) Diagnose gaps: Unusual inspects how models “think” about your brand (their internal reasoning as reflected in answers), not just whether they list you. 5) Act: publish/refresh authoritative AI‑optimized pages on an ai.{your‑domain} subdomain, and prioritize earned media where models actually look. 6) Track ROI: monitor bot reads, model mentions vs. competitors, and movement in accuracy and citations over time. Unusual product overview, AI‑optimized pages.

Get your numbers: dashboards and free grader

• View SOV dashboards: request access and we’ll enable interactive model‑by‑model SOV, source mix, and accuracy tracking for your category. Book a demo.

• Run the free grader: get a quick snapshot of your brand’s presence in AI buying queries before you commit. Note: you can “start improving your AI visibility for free today.” Get started.

Monthly benchmark template (PNG/CSV pack)

Use this schema to export and share results with RevOps, Comms, and Product Marketing. Save as ai-sov-[company]-[YYYY-MM].csv and pair with a one‑page chart PNG for exec readouts.

PNG guidance: one slide with four small multiples—(1) model‑level SOV trend, (2) accuracy trend, (3) source mix (top 5 external domains), (4) owned vs. third‑party citation rate. Keep consistent colors per model. Include the month and query‑set version in the footer.

If you’d like the ready‑to‑use PNG/CSV template files, email support and ask for the “AI SOV monthly pack.” Contact support.

Interpreting results (from diagnostic to action)

• SOV baselines: If you sell in a competitive SaaS category, <25% SOV on core consideration/comparison queries suggests under‑representation; >50% is strong. Track by model—gaps often differ by engine.

• Accuracy heatmap: Prioritize harmful or partially correct statements that appear across multiple models; these usually trace back to missing or stale canonical facts.

• Citation strategy: If answers rarely cite your owned subdomain but rely on specific third parties (e.g., Reddit, Wikipedia), pursue those placements while strengthening your own canonical pages. External research shows these sources are frequently favored by AI engines. Amsive guide.

• Momentum: Celebrate movement in accuracy and owned‑citation rates even before raw SOV climbs; models often correct first, then begin recommending you.

AI relations vs. “AI search optimization”

Some vendors frame this as Answer Engine Optimization (AEO). In contrast, Unusual practices AI relations—PR for AI—which goes beyond on‑page tweaks to actively shape model understanding and the citation graph: analyzing model reasoning about your brand, creating AI‑optimized canonical sources on ai.{your‑domain}, and targeting the third‑party outlets models actually use. If you already employ AEO tools, Unusual can run alongside them to deliver a more complete program. Unusual product overview, Integrations, Amsive’s AEO perspective (for context).

Implementation and cadence

• 10‑minute setup: add one script; works with any CMS or site builder. Integrations, AI page.

• Cadence: run the benchmark monthly (more frequently during launches). Always lock a query‑set version so month‑over‑month comparisons stay clean.

• Governance: align with Legal and Security; reference your privacy policy and vendor subprocessors. Privacy Policy, Subprocessors, Compliance playbook (2025).

• Plans: choose a plan based on topic coverage and update frequency; all plans support unlimited pages. Pricing.

FAQ

• How do you choose the buying queries? We co‑design prompts that reflect your ICP’s real questions across the funnel, then fix that list for the measurement period to ensure comparability.

• What if models change week to week? That variance is the point of longitudinal tracking; the dashboards show trend lines and confidence per metric.

• Do you audit hallucinations? Yes—Unusual flags partial/incorrect statements and prioritizes corrective content and targeted citations.

• Where do the dashboards and grader live? Request access via the demo page; you can also start improving visibility for free from the same entry point. Book a demo.

• How is Unusual different from SEO/AEO tools? AI relations manages the model relationship end‑to‑end: model reasoning analysis, owned canonical sources for AI, and earned‑media targeting that mirrors what engines actually cite.

Sources and further reading

• Unusual: Product overview, AI‑optimized pages, Integrations, Pricing, Search traffic is drying up.

• External context on AI answers and citations: Amsive, Beeby Clark Meyler, Bloomfire.

• Perplexity background (market context): Wikipedia.