AEO EngineAnswer Index correction log
Rankings are never for sale. Client firms are always disclosed. Corrections within 48 hours.
Every correction to a published edition is logged here with the edition, firm, field, old value, new value, date and reason. Report an error to team@aeoengine.ai; corrections are investigated and resolved within 48 hours.
| Date | Edition | Entry |
|---|---|---|
| 2026-09-08 | All editions | Correction: Firms measured at zero mentions and therefore tied at 0.0% are now unranked, and are listed alphabetically in a separate cohort instead of carrying a numbered position. The August 2026 files gave every firm an ordinal, so a firm with no measured mentions still read as a ranked position it had not earned and which reshuffled arbitrarily between editions. |
| 2026-09-08 | All editions | Correction: the note explaining why month-over-month change is not shown has been replaced. It had been written for internal use and named the specific vendor model builds behind the ChatGPT column, along with internal pipeline states and raw record counts, none of which is part of the published methodology. Every edition now reads "Month-over-month change is not shown for this edition." No ranking, share of voice, appearance rate, engine flag or firm row was changed: the four September data files differ from the versions published earlier on 2026-09-08 in that one sentence and in nothing else. The engines are still named as ChatGPT, Gemini, Perplexity and Google AI Overviews throughout, and the measurement methodology is unchanged. |
| 2026-09-08 | All editions | Correction: Updated the concentration summary so the leading-cohort and remaining shares are added up from the same one-decimal percentages published in the board and the CSV. Previously the summary summed full-precision values and rounded once at the end, which could differ by a tenth from the figures a reader adds up: the September National edition read 51.3% where its own published column totals 51.2%. Both halves of the sentence are now summed from the published column rather than one being inferred from the other, and they still total 100.0%. |
| 2026-09-08 | All editions | Methodology clarification: from the September 2026 edition, share of voice is defined as a firm's share of all AI mentions measured in its edition, so the column sums to 100%. The August 2026 files were assembled by hand and their share column coincidentally summed to about 100%; no August number has been changed. Appearance rate is a separate published column with its own denominator: the share of the edition's frozen questions in which the firm appeared at least once, and it does not sum to 100%. |
| 2026-09-08 | All editions | Methodology clarification: the published share-of-voice column is rounded to one decimal by largest-remainder apportionment, so each edition totals exactly 100.0% and the board, the CSV, the press kit, the badges and the concentration summary all quote the same figure. Rounding each firm independently does not preserve the total: the September editions would otherwise have totalled 99.8%, 100.5%, 100.0% and 100.5%. A published figure can therefore differ from the underlying share by up to 0.05 of a percentage point. Ranking order is unaffected, and any firm with at least one measured mention publishes at least 0.1%, so 0.0% continues to mean exactly what it says: no measured mention. |
| 2026-09-08 | National 100 | Correction: badge URLs for the National 100 edition resolve again. The embed code shown in the brag kit briefly pointed at a badge path that did not exist for this edition; the published path is unchanged from August (/answer-index/badges/national-100-TICKER.png) and badges already embedded on firm websites were never affected. |
| 2026-09-08 | National 100 | Correction: the August 2026 National board was a provisional edition, not a measured baseline. September 2026 is the first measured National edition. Its 100-firm roster is retained for subsequent monthly editions, and National movers and decliners start with October 2026, compared with September. The August board remains available as originally published in the archive; its figures are not used to calculate month-over-month change. No September ranking or score has changed. |
| 2026-09-08 | Personal Injury 100 | Correction: 26 firms that appeared in the August 2026 Personal Injury edition were missing from the September eligibility list and were restored before publication, returning the edition to its full 100-firm universe. Their September mentions were extracted from the already-collected September answers; no new measurement was run and no other edition was affected. Affected firms include Gair, Gair, Conason, Wilshire Law Firm, Kline & Specter, Beasley Allen, Cellino Law, Zehl & Associates, Weitz & Luxenberg and Motley Rice. |