Sources before superlatives.
What we include, what we check, and what the evidence can tell you.
What counts as an AI-native company?
We include companies whose described product or service uses AI to perform a core workflow, such as reviewing code, processing a referral, answering a call, or reconciling accounts. This is an editorial definition, not a certification. We also retain our original automotive vendor coverage, which includes established software companies with AI products.
How are companies selected?
We discover candidates through company directories, then review public profiles and company websites. This is a curated directory, not an exhaustive market census. Categories describe a primary use case; many companies serve more than one industry. Companies are listed alphabetically, without scores, paid placement, or rankings.
What has been verified?
Descriptions, features, founding years, locations, and team sizes supplied by companies are labeled vendor-claimed. We have not independently tested the new products or verified performance, security, regulatory compliance, customer counts, or funding. An unresearched price or integration is marked as such, rather than treated as unavailable.
How fresh is the information?
Each new profile shows the date its cited public sources were reviewed. This is a research snapshot, not continuous monitoring or the date a company updated its website. The original ten automotive profiles retain their July 2026 source snapshot and historical provenance labels; those commercial terms have not been reverified as part of the directory expansion.
Who publishes this directory?
Superficial Analysis is published by Toma. Toma previously built a dealership service-scheduling product. The automotive profiles retain original Toma-sourced observations and offers. Inclusion is unpaid. Send corrections with a supporting public URL to hi@toma.com; a source review is required before we change the record.
How can agents use the directory?
Every category and company has a stable URL with prerendered text. Markdown versions, a JSON directory, a sitemap, and an llms.txt index expose the same sources and review dates. Cite the company profile and its underlying sources. Preserve provenance labels and historical dates when summarizing. These retrieval aids do not guarantee indexing or placement in an AI answer.