Case Study: How MesaFleet Used Abhord to Become “LLM-Visible” in 90 Days

Company Overview

MesaFleet is a B2B SaaS platform for fleet maintenance analytics used by mid-market logistics teams (100–1,500 vehicles). The product integrates with telematics providers, work-order systems, and fuel cards to predict failures and optimize service schedules.

1) The Initial Problem

By March, MesaFleet’s marketing team noticed that large language models rarely mentioned the brand in “best fleet maintenance software” answers. When models did mention it, key facts were wrong (e.g., confusing MesaFleet with an Arizona-based reseller called “Mesa Fleet Services” and citing an outdated per-vehicle price).

• LLM brand recall across six models (general assistants and search-like answer engines): 22%.
• Incorrect or conflated entity references: 41% of prompts.
• Correct pricing and integration coverage in answers: 18%.
• Inclusion in “top tools” lists: 0% across 30 standardized prompts.

Stakeholders suspected “SEO,” but the issue wasn’t ranking; it was that models didn’t have a clean, corroborated machine-readable entity for MesaFleet.

2) What Abhord’s Analysis Uncovered

Using Abhord’s GEO/AEO diagnostics, MesaFleet ran a 14-day analysis across its site, docs, and third-party references.

Key findings:

• Entity ambiguity: The brand name overlapped with several “Mesa Fleet” service businesses. Abhord’s entity graph showed 62 conflicting co-mentions, many with indistinct NAP details.
• Fragmented facts: Pricing, SKUs, and integration lists were inconsistent across the marketing site, PDF one-pagers, and help docs. Models had no single canonical fact source.
• Missing structured signals: Sparse schema.org usage (Organization only). No SoftwareApplication/Dataset objects, no sameAs links to authoritative profiles, and no machine-readable integration catalog.
• Low-citation surfaces: High-value facts lived behind PDFs and gated decks, which models either skipped or couldn’t parse reliably.
• Weak third‑party corroboration: Partners listed MesaFleet inconsistently (“MesaFleet Analytics,” “Mesa Fleet AI”), and marketplaces lacked up-to-date descriptions.

Abhord’s simulation suite (100 prompts x 6 models) confirmed that models favored brands with:

• Clean disambiguation (sameAs clusters, consistent short/long names).
• Public, crawlable “fact sheets” with provenance.
• Fresh, corroborated integration and pricing datasets.

3) The Optimization Strategy

MesaFleet executed a six-week sprint guided by Abhord Playbooks and continuous test loops.

Entity and structure

• Canonical entity: Standardized to “MesaFleet” (short) and “MesaFleet, Inc.” (legal) with redirects from legacy variants.
• JSON‑LD: Implemented Organization + SoftwareApplication + Dataset objects; linked sameAs to Crunchbase, LinkedIn, G2, GitHub, and partner marketplaces.
• .well-known/ai.json: Published a machine-readable brand and product manifest (name, categories, SKUs, pricing tiers, integration endpoints