AI Visibility Benchmark · Dataset
Service-Business AI Visibility Benchmark Dataset
This page hosts the reviewed public dataset for the SEO Informatica AI Visibility Benchmark 2026.06.
Benchmark record
- Version
- 2026.06
- Run ID
- semantic-reviewed-50-2026-06-04
- Records
- 50 reviewed records
The dataset is designed to support benchmark claims about public service-business page readiness. It is not designed to expose the audited domains, publish private company findings, or claim market-wide averages.
Dataset Summary
| Field | Value |
|---|---|
| Dataset name | SEO Informatica AI Visibility Benchmark Dataset |
| Version | 2026.06 |
| Run ID | semantic-reviewed-50-2026-06-04 |
| Sample size | 50 reviewed records |
| Distinct retained site IDs | 50 |
| Collection period | June 3, 2026 UTC / June 4, 2026 IST |
| Public export revision | October 5, 2026: direct identifiers removed; June observations and scores retained |
| Domain publication | Direct URLs, page text and linkable hashes excluded; stable site IDs retained |
| Reuse enquiries | Contact us about reuse permissions. The citation below identifies the source; it is not a licence grant. |
| Reviewed CSV download | /downloads/ai-visibility-benchmark-2026-06-reviewed.csv |
| Reviewed JSON download | /downloads/ai-visibility-benchmark-2026-06-reviewed.json |
| Codebook | /downloads/ai-visibility-benchmark-codebook-2026-06.md |
| Rubric | /downloads/ducr-scoring-rubric-2026-06.json |
What Is Public
The corrected export contains 50 rows and 80 fields: stable site IDs, recorded categories, typed page/access/schema observations, stored DUCR scores, review status and dated run identifiers. It retains every value in those included columns, including blanks.
The earlier export retained identifying URLs and page text despite its anonymization description. The October 5 correction removes 26 fields from the reviewed CSV/JSON and retains nine non-identifying fields in the override CSV. Direct URLs, titles, descriptions, H1 text, freeform notes, location fields and linkable domain/source hashes are excluded. Stable IDs can still be linked to earlier downloaded versions; this correction does not make previous copies anonymous.
No new crawl or rescoring was performed. The original study edition, 50 site IDs, retained observations, category scores and summary statistics are unchanged. The codebook lists the exact retained fields and missing-value rules. The checksum manifest describes the current public files, including the separately updated monthly tracking template.
Field Groups
- Site and sample metadata
- Direct URL and identity fields omitted from the corrected export
- Crawl access fields
- Index and snippet fields
- Page-structure fields
- Entity and source-clarity fields
- Schema fields
- DUCR scoring fields
- Manual semantic review status
- Provenance fields
Review Status
| Review Status | Count |
|---|---|
semantic_review_approved |
22 |
semantic_review_approved_with_caveat |
28 |
Keep caveat labels when quoting or analysing a row. Blank fields mean unrecorded or unspecified, not false. In particular, all author-name and modified-date fields are blank. The public file does not provide a complete per-criterion point ledger or original page HTML.
Downloads
- Reviewed CSV: /downloads/ai-visibility-benchmark-2026-06-reviewed.csv
- Reviewed JSON: /downloads/ai-visibility-benchmark-2026-06-reviewed.json
- Summary JSON: /downloads/ai-visibility-benchmark-summary-2026-06.json
- Benchmark stats JSON: /downloads/benchmark-stats-2026-06.json
- Codebook: /downloads/ai-visibility-benchmark-codebook-2026-06.md
- DUCR rubric JSON: /downloads/ducr-scoring-rubric-2026-06.json
- Semantic review overrides: /downloads/semantic-review-overrides-2026-06-04.csv
- SHA256 checksums: /downloads/SHA256SUMS.txt
How To Cite The Dataset
Use the dataset version and date, not a vague page title.
Recommended citation format:
SEO Informatica. “Service-Business AI Visibility Benchmark Dataset.” Version 2026.06. Collected June 3, 2026 UTC / June 4, 2026 IST. Reviewed sample of 50 service-business websites. Corrected public export, October 5, 2026.
Known Gaps
This dataset uses public-page audits. It cannot measure private analytics, CRM outcomes, hidden platform behavior, personalized AI answers, or guaranteed citation probability.
The sample is vertically uneven: consulting, accounting and agency sites represent 35 of 50 rows. All original country fields were coded US. Category comparisons are descriptive and do not establish wider market differences. Accessible public sites were selected, which also limits conclusions about access problems elsewhere.
Related Pages
For a practical interpretation, read when a finding warrants work or discuss your business question.
The Outside View
See where you stand with Google, AI and your buyers.
Send us your website. In two business days you have it in writing: how you look from the outside, who really competes with you, and the ten fixes that matter. Free.