We Audited a Local Garage 5 Times in 10 Days. Here's What AI Search Actually Says About It.

A real ASEO audit baseline — before a single change was made to the website.

The hardest thing to sell in AI search optimisation is proof. Anyone can promise improved visibility. What the market doesn't have yet is documented, before-and-after audit data from a real business — not a demo, not a simulation, not a hypothetical. Real prompts, run across five AI platforms, on a business that had no idea what its AI citation score was.

Between 15 March and 25 March 2026, Cited By AI® ran five consecutive ASEO audits on an independent Audi and Volkswagen specialist in Alberta, Canada. Fifteen years in business, strong local reputation, zero AI visibility strategy. No changes were made to the website throughout the audit period. This is the baseline.

The business profile

Independent Audi and VW specialist. Fifteen years trading in a mid-sized Canadian city. Strong word-of-mouth, established Google presence, and a website that reads well to humans. The kind of business traditional SEO would consider healthy.

What AI search says about it is a different story.

The CPS® baseline

The audit analysed 29 content blocks across 14 pages in the most comprehensive run. Overall CPS® came in at 52.7 — Grade C, Occasionally Cited. AI platforms surface this business inconsistently: sometimes it appears, sometimes it doesn't, depending on how the query is phrased and which platform is asked.

Breaking it down by pillar shows exactly why.

Fact Density at 23.6 is the standout strength. The site content contains strong concentrations of specific facts, named entities, and verifiable data. AI systems value specificity and this content has it. This pillar alone is why the business appears in AI answers at all.

Content Structure at 10.8 is the critical weakness. Content blocks aren't in the 134–167 word optimal RAG chunk range — either too short to provide sufficient context, or too long, diluting the key facts. This single pillar accounts for the majority of missed citations.

Answer Structure at 6.6. Every content block opens with brand narrative rather than a direct declarative answer. The difference between those two approaches is concrete:

The difference in citation probability between those two opening sentences is measurable. The first gives an AI system nothing to retrieve. The second gives it five verifiable signals in 22 words.

Freshness Signals at 1.0. Almost no date markers or recency language anywhere on the site. For Perplexity — which already delivers the highest Share of Voice for this business at 46.7% — freshness is a primary citation factor. This is the easiest fix available and the highest per-effort return on the entire site.

Self-Containment at 10.6 is average. Some blocks read well in isolation; others contain dangling references that require surrounding context. AI retrieval extracts blocks without their surrounding paragraphs. Context-dependent content doesn't get cited.

What AI platforms actually see

Across five audit runs, the business appeared in 23.5% to 32.7% of all prompts. Perplexity consistently led at 46.7% to 62.7% Share of Voice — reflecting its known behaviour of heavy off-page citation, where the business's existing local presence is strong enough to drive results. ChatGPT, Gemini, Claude, and Copilot all showed similar performance in the 13–30% range.

The funnel stage picture is where the business opportunity becomes clear. Decision-stage visibility is solid — buyers who already know the business find it in AI answers. Awareness and Consideration stage visibility is the gap. Buyers who don't yet know this specialist exists, or who are comparing options across local mechanics, aren't seeing it.

The competitor landscape

The competitor picture in AI search for this location and category is fragmented. No single competitor dominates consistently across all five audit runs. More significantly, 35 of 49 total competitor appearances in the final run were "none mentioned" — AI systems frequently respond to location-specific automotive queries without recommending any specific business at all.

The five priority fixes

Based on five audit runs, these are the highest-ROI changes available — all content and markup, no infrastructure changes required.

• 1
  Fix Content Structure first

  Restructure all content pages so individual sections contain 134–167 words. This addresses the highest-weight CPS® pillar and delivers the largest score improvement of any single change.

• 2
  Fix Answer Structure second

  Replace all narrative-style content openings with direct declarative patterns. Every section should open with a sentence that answers the most likely question a buyer would ask.

• 3
  Add Freshness Signals third

  Add "As of 2026" language and date markers throughout the site. Fifteen minutes of work on a site with near-zero freshness signals will move this pillar from 1.0 to near-full marks — the highest per-effort return available.

• 4
  Deploy AutoRepair schema with areaServed and knowsAbout fields

  The site currently uses generic LocalBusiness schema. AutoRepair schema with explicit location fields and specialist capability declarations makes the business eligible for queries it currently doesn't appear in at all.

• 5
  Add FAQ sections to all service pages

  Five to eight questions per page, phrased exactly as buyers ask AI systems. FAQ content is cited at significantly higher rates than standard prose because the question-and-answer format mirrors how AI systems construct their responses.

Expected outcomes from full implementation

The follow-up audit runs after optimisations are applied. Before/after data will be published when available.