Part of the AEO / GEO / AI-search audit tool stack. See the pillar post for the full catalog of sibling audits and where this one fits in the lineup.
The Semrush piece on the agentic web landed on a point I'd been circling without writing down: the old SEO playbook rewarded generic positioning. "CRM for businesses" matches more queries than "CRM for automotive dealerships" — so for decades, marketers chose the broader framing and let the SERP sort out who actually fit.
Agents reverse this. When an automotive buyer's procurement agent runs a selection query, it's already decided the buyer is an automotive dealership. The question becomes: which of the candidate CRMs explicitly says "for automotive dealerships"? The generic page drops out of consideration not because it isn't good, but because the agent can't verify fit from the marketing copy.
So the thing that used to be a liability — positioning narrowly enough to exclude most visitors — becomes the thing that gets you selected. And the thing that used to be smart — broad positioning so everyone can imagine themselves in the customer photo — becomes the reason you don't show up in agentic buying at all.
The tool at /tools/ideal-customer-declaration-audit/ audits whether your product page declares its ideal customer with the specificity an agent can match against. It checks six things.
Does the H1 or title name an industry? "Automotive dealership CRM" beats "Cloud-based CRM software" for any agent running an automotive selection query. Does the body declare target company size — SMB, mid-market, or better, an employee range like "10-50 seats"? Does it name a geography? Does it use "built for X who need Y" patterns anywhere? How many industries and use-cases are named in total? And does any of it show up as structured audience data in JSON-LD, or is it all natural language the agent has to interpret?
The scorecard grades each signal separately and emits a fix prompt with a ready-to-paste Audience schema block. The schema.org audience type is genuinely easy — audienceType plus geographicArea. Most pages skip it entirely, and adding it is a one-line win most competitors won't bother with for another six months.
The weirder piece of advice in the fix prompt: add a "not a fit for" section. Three or four bullets listing the kinds of customers this isn't for. This reads like it should cost you reach, and in SERP-land it did. In agent-land it raises the confidence of every positive match — the agent sees an explicit disclaimer and treats the positive fit as more trustworthy, not less. Pages that try to be everything to everyone lose to pages that commit.
What this tool can't do is decide the ICP for you. If your product genuinely serves everyone from solopreneurs to Fortune 500s on the same feature set at the same price point, no amount of audit output fixes that — the problem is the pricing or the product, and the positioning is just downstream of those decisions. The audit will flag it as low, but the fix is strategic, not a copy rewrite.
For the product-page specifics beyond audience declaration — schema completeness, Merchant Center readiness, checkout friction — pair this with the Agentic Commerce Readiness Audit. The two together cover the page-layer half of agent-selection eligibility.
If the agentic web is where you're betting your product discovery, the end-to-end playbook sits inside The $20 Dollar Agency — it's about building repeatable client-ready service offerings rather than software, but the ICP-declaration logic is the same.
Related reading
- AI-Citation Specificity Audit — paired gap-fill tool
- Pillar-Cluster Topology Audit — paired gap-fill tool
- Agentic Commerce Readiness Audit
- Cross-Domain Entity Consistency
Fact-check notes and sources
- Agentic-web positioning argument: Semrush — The Agentic Web.
- schema.org audience type reference: schema.org/Audience and schema.org/audience property.
- Microsoft AI Max + agent selection model: Search Engine Land — Microsoft launches AI Max.
- Model Context Protocol: modelcontextprotocol.io.
Informational, not marketing or positioning advice. Mentions of Semrush, Microsoft, Anthropic, and similar products are nominative fair use. No affiliation is implied.
