The 2026 State of AEO: 1,168 Sites Across 11 Industries
Abstract. We audited 1,168 websites across 11 industries between January and April 2026 using a 5-category AEO scoring system calibrated against published research from SE Ranking, Ahrefs, and Previsible. The mean AEO score across the corpus was 60.6/100. 14% of sites scored below 50 (effectively invisible to AI search), 46% scored below the 65 baseline for AI citation, and zero sites in the corpus scored 90+. Trade-services sites (HVAC, electrical, plumbing) averaged 56–62. DTC ecommerce averaged 58.9. Law firms scored highest at 67.3 — yet still none reached top-tier citation thresholds. Most of the missing AEO points came from absent schema markup, missing E-E-A-T signals, and content not structured in question-and-answer form.
Why this matters
Google's AI Overviews now appear on 47% of searches (Google Search Labs, 2026). ChatGPT, Perplexity, and Bing Copilot capture buyer intent before users see organic results. The brands those AI engines cite earn the click. The brands they don't cite are invisible — even when their domain ranks well in traditional SEO.
Yet the SEO industry still treats AEO as a side conversation. There are no published, primary-source benchmarks for AEO scores by industry. So we built one — by running our own audit engine across 1,168 real production websites and publishing what we found.
Headline findings
Five things that surprised us
- Zero sites scored 90+. Across all 1,168 audits, not a single site had crossed into top-tier AI-citation territory. The "best" sites are merely citable (75+) — not preferred sources.
- Trade-services lag 4–10 points behind professional services. Appliance repair averaged 52.5; law firms averaged 67.3. The gap is almost entirely E-E-A-T (named author bylines, credentials, original content) — which trades sites systematically lack.
- HVAC has the largest pool, the median performance, and the biggest tail. 421 HVAC sites in our corpus, averaging 62.3, ranging from 21 to 81. Variance within the niche is wider than across niches.
- Schema markup is the single weakest category. Across the corpus, 60% of sites have no JSON-LD at all. The remaining 40% mostly have only
Organizationschema — none of the high-leverage types likeFAQPage,HowTo, orArticle. - The threshold for AI citation is binary, not gradual. Sites scoring 50–64 (46% of the corpus) are essentially uncited even when their domain authority is strong. Crossing 65 produces a meaningful jump in observed citation rate, with the cleanest outcomes at 75+.
AEO score distribution across 1,168 audits
| Score band | Interpretation | Sites | % of corpus |
|---|---|---|---|
| 0–29 | Critical: AI cannot read or cite | 14 | 1.2% |
| 30–49 | Poor: technically reachable but rarely extracted | 151 | 12.9% |
| 50–64 | Below baseline: occasional mentions, no citation primacy | 537 | 46.0% |
| 65–74 | Approaching citation: sometimes surfaced, never first | 390 | 33.4% |
| 75–89 | Citable: regularly extracted on relevant queries | 76 | 6.5% |
| 90–100 | Top-tier: AI's first source on the topic | 0 | 0.0% |
AEO scores by industry
Sorted from worst to best average. n = sample size in that niche. %<50 = share of sites scoring below 50 (invisible to AI). %75+ = share at AI-citable threshold or above.
| Industry | n | Avg | Min | Max | % < 50 | % ≥ 75 |
|---|---|---|---|---|---|---|
| Appliance repair | 22 | 52.5 | 17 | 78 | 31.8% | 4.5% |
| Electricians | 103 | 55.7 | 18 | 83 | 26.2% | 4.9% |
| Veterinarians | 109 | 58.4 | 30 | 85 | 16.5% | 3.7% |
| DTC ecommerce brands | 154 | 58.9 | 21 | 79 | 18.8% | 3.9% |
| Plumbers | 84 | 59.7 | 28 | 78 | 17.9% | 4.8% |
| Pool installers | 16 | 60.3 | 39 | 78 | 25.0% | 12.5% |
| Roofing contractors | 122 | 61.4 | 25 | 81 | 11.5% | 3.3% |
| Dental practices | 68 | 61.9 | 44 | 79 | 11.8% | 11.8% |
| HVAC contractors | 421 | 62.3 | 21 | 81 | 9.7% | 8.1% |
| Garage door repair | 21 | 63.4 | 44 | 78 | 9.5% | 9.5% |
| Law firms | 46 | 67.3 | 55 | 81 | 0.0% | 10.9% |
What the lowest-scoring sites consistently miss
We tagged the most common issues across the bottom-quartile sites. Five issues account for over 70% of the missing points:
- No FAQPage schema. 89% of audited sites have no question-answer schema markup. AI engines extract directly from this format — its absence is the single largest extractable-content gap.
- Robots.txt blocking AI crawlers. 23% of sites in our corpus block at least one of GPTBot, ClaudeBot, CCBot, or PerplexityBot — often inadvertently, as a side effect of legacy SEO plugins.
- No named author bylines. 78% of content pages have no human author. AI engines lean heavily on E-E-A-T signals, and an unnamed source ranks below a named one.
- No publish or update date visible on the page. Dated content gets cited at materially higher rates than undated content — yet 64% of audited pages have no visible date.
- Content structured as paragraphs, not Q&A. Even when sites discuss buyer-relevant topics, the prose format is hard to extract. Question-headed sections drive a 180% lift in citation rate (Previsible, 2024).
Methodology
Our scoring system breaks AEO into five weighted categories, calibrated against published research:
- Content Structure (30% weight) — question-headed sections, list density, paragraph clarity, internal heading hierarchy
- Technical & Crawlability (25%) — robots.txt allowances for AI bots, Core Web Vitals, response time, internal linking
- E-E-A-T Signals (23%) — named author with credentials, original research presence, citations from external sources, contact information
- Schema Markup (12%) — JSON-LD types present, validated structure, primary-entity coverage
- Meta & Technical (10%) — title hygiene, meta descriptions, Open Graph, canonical tags
Each site was crawled to a maximum of 20 pages. Per-page scores were averaged at the site level. Audits were run between January 1, 2026 and April 27, 2026. The full scoring rubric is available on request.
How to cite this research
Published under CC BY 4.0. Quote, cite, or republish freely with attribution.