# AI Visibility Benchmark 2026.06: corrected public-export codebook Study edition: 2026.06. Collection: June 3, 2026 UTC / June 4, 2026 IST. Public-export correction: October 5, 2026. Fifty retained rows; 80 fields. ## What changed The earlier export retained identifying URLs and page text despite its anonymization description. This correction removes 26 columns from the reviewed CSV/JSON and retains nine non-identifying columns in the override CSV. Direct URLs, titles, descriptions, H1 text, freeform notes, location fields and domain/source hashes are excluded. Stable site IDs remain for record analysis and can be linked to earlier downloaded versions. This is not a promise of irreversible anonymity, nor does it remove copies already obtained elsewhere. The June dates, row order, site IDs, included observations, category scores and aggregate results are retained without rescoring. The original records and previous file hashes are preserved in the research archive. No new crawl or sample was introduced. The original June rubric and summary/stat files retain their bytes. ## Reading values CSV and JSON use the same field names, order and scalar strings. Counts and scores are numeric strings, flags are the strings `true` and `false`, and blank strings mean unrecorded or unspecified. A blank does not mean false, zero or confirmed absence. The CSV has one header and 50 records; JSON is an array of the same 50 objects. Neither format changes the historic type representation. All author-name, published-date and modified-date fields are blank in this edition. Original-data flags, by contrast, explicitly record false in all 50 rows. The earlier overview described author/date counts as zero; blank fields do not establish absence. Treat detector flags and counts as stored observations, not independent current-page verification. The robots file response was 200 for 48 rows, 406 for one and 404 for one. Both non-200 rows retain true allowed flags for all ten tokens. The package does not supply the collector's handling of unsuccessful robots responses. These flags are not authenticated platform visits or confirmed indexing/citation outcomes. HTTP service responses are those obtained by the study collector. Sitemap discovery was bounded in the documented June protocol. Negative discovery flags do not prove a page was absent from every sitemap. All original country values were coded US; country and region are omitted from this corrected row-level export. Category comparisons describe this selected sample, not wider markets. ## Scores and review DUCR category maxima are Discoverable 25, Understandable 25, Citable 30 and Routable 20. The four stored values sum to the stored total in all 50 rows. Twenty-two records are approved and 28 approved with caveat. Keep review status attached to any selected example. The historical rubric supplies prose criteria and weights, but the public files do not include individual criterion awards, the complete scoring implementation or underlying pages. You can reproduce record counts, totals, means, medians and distributions. You cannot independently replay the historical audit or derive every category score from the public fields. The rubric includes benchmark- and SEOI-specific assumptions; it is not a universal platform quality standard. No citation, traffic, enquiry or sales outcomes were measured in this study. ## Retained dataset fields - `run_id` - `record_id` - `audit_date` - `collector_version` - `anonymized_site_id` - `domain_public` - `vertical` - `http_status_home` - `http_status_service` - `robots_fetch_status` - `robots_allows_googlebot` - `robots_allows_bingbot` - `robots_allows_oai_searchbot` - `robots_allows_gptbot` - `robots_allows_chatgpt_user` - `robots_allows_claudebot` - `robots_allows_claude_searchbot` - `robots_allows_claude_user` - `robots_allows_perplexitybot` - `robots_allows_perplexity_user` - `waf_block_signal` - `captcha_signal` - `js_challenge_signal` - `sitemap_found` - `sitemap_url_count` - `key_pages_in_sitemap` - `noindex_present` - `nosnippet_present` - `max_snippet_value` - `data_nosnippet_on_core_content` - `canonical_self_service` - `visible_word_count_service` - `h1_count_service` - `h2_count_service` - `h3_count_service` - `answer_block_count` - `table_count` - `list_count` - `faq_visible_count` - `section_anchor_count` - `hidden_core_content_flag` - `pdf_core_info_flag` - `image_only_core_info_flag` - `rendered_text_ratio` - `org_name_present` - `service_name_consistent` - `location_info_present` - `about_page_present` - `contact_details_present` - `author_name_present` - `date_published` - `date_modified` - `version_history_present` - `sameas_count` - `external_citation_count` - `official_source_citation_count` - `original_data_present` - `methodology_present` - `limitations_present` - `dataset_download_present` - `proof_examples_count` - `claim_support_score` - `schema_types_detected` - `organization_schema_present` - `person_schema_present` - `webpage_schema_present` - `article_schema_present` - `dataset_schema_present` - `faqpage_schema_present` - `breadcrumb_schema_present` - `howto_schema_present` - `schema_visible_content_match` - `ducr_discoverable_score` - `ducr_understandable_score` - `ducr_citable_score` - `ducr_routable_score` - `ducr_total_score` - `critical_blocker` - `manual_review_status` - `crawl_timestamp` ## Removed fields - `auditor_id` - `sample_source` - `domain_hash` - `site_url` - `country` - `region` - `business_type` - `lead_gen_type` - `homepage_url` - `primary_service_url` - `support_article_url` - `about_url` - `contact_url` - `location_url` - `robots_url` - `meta_robots_service` - `x_robots_tag_service` - `canonical_url_service` - `page_title_service` - `meta_description_service` - `h1_service` - `author_bio_url` - `reviewer_name` - `reviewer_notes` - `source_html_hash` - `crawl_user_agent` ## Retained override fields - `anonymized_site_id` - `manual_review_status` - `org_name_present` - `service_name_consistent` - `proof_examples_count` - `claim_support_score` - `original_data_present` - `methodology_present` - `limitations_present` ## Provenance and checksums Use the study edition and the export-correction date when citing these files. The public SHA256SUMS.txt describes the current nine listed files. It also accounts for the separately revised monthly citation template, whose bytes are preserved in this correction. A checksum verifies file bytes, not a claim's truth or the anonymity of historical copies. For reuse permissions, contact SEO Informatica. The citation convention is not a licence grant.