AEO Scorer, Benchmark and Methodology
The rubric we score against, plus the public corpus of brand blog posts we used to calibrate it. Everything is reproducible.
TL;DR: We scored 50 brand blog posts against a 6-dimension, 100-point rubric built around how Google AIO, ChatGPT, and Perplexity decide what to cite. The corpus average is 44.4/100, and 70 is the threshold where citation probability rises meaningfully.
How do we score across 6 dimensions for 100 points?
Each post is graded against six dimensions that map to how Google AIO, ChatGPT, Perplexity, and Bing actually pick what they cite.
| Dimension | Weight | What it checks |
|---|---|---|
| Authority and E-E-A-T | 30 pts | Visible byline, external citations, original data, and JSON-LD structured markup that proves who wrote it. |
| Entity Specificity | 20 pts | Named entities, numeric depth, and concrete citations rather than vague nouns. |
| Structural Extractability | 20 pts | TL;DR up top, scannable bullets, real tables, and schema that a model can lift cleanly. |
| Q-Subheadings | 10 pts | Question-form headings that match the actual queries users type. |
| Semantic Clarity | 10 pts | Answer-first openers, low jargon, no hedging language. |
| Freshness | 10 pts | Recency signals, updated timestamps, and current-year statistics. |
Anti-gaming cap. If a draft trips 2 or more stuffing patterns, the total is capped at 60 regardless of dimension scores.
Letter grades. A is 90 and above, B is 80 to 89, C is 70 to 79, D is 60 to 69, F is below 60.
A seventh dimension, Voice and Brand Match, exists in the internal skill but is dropped for public-tool comparison so scores stay apples to apples.
Why does this rubric predict citation probability?
Large language models pull from pages they can parse confidently, attribute clearly, and present to a user without embarrassment. Each dimension maps directly to one of those three requirements.
Authority and E-E-A-T (30 pts) is the highest-weighted dimension because AI engines have to justify their citations. A visible byline, external links to authoritative sources, original data, and JSON-LD structured markup all serve as machine-readable proof of credibility. Pages that omit even one of these signals drop significantly in retrieval ranking.
Entity Specificity (20 pts) separates pages that mention "AI tools" from pages that name specific products, include version numbers, and cite named studies. Vague nouns are hard for a model to anchor. Concrete named entities are easy to extract and quote.
Structural Extractability (20 pts) is what makes a page citable rather than merely findable. A TL;DR at the top gives a model a ready-made answer. Scannable bullets and real tables let it lift structured data without paraphrasing. Schema markup confirms the page type, author, and publication date in a format every major AI crawler reads.
Q-Subheadings (10 pts) align page structure with the queries users actually type. When a heading matches a natural-language question, the content beneath it becomes a candidate answer for that specific query. Pages with zero question-form headings lose this signal entirely, which is why so many otherwise solid posts score in the 30-to-50 range.
Semantic Clarity (10 pts) penalizes hedging language, unexplained jargon, and paragraph-first openers. AI engines prefer pages that lead with the answer and explain the reasoning after, not before. Every "it depends" and every buried conclusion is a point lost.
Freshness (10 pts) rewards recency signals: updated timestamps, current-year statistics, and in-text references to recent events. A page that was excellent in 2023 but has no freshness indicators scores lower than a comparable 2026 post because AI engines actively filter for up-to-date information on fast-moving topics.
The 70-point threshold we publish as the citation target comes from calibrating this rubric against the 50-entry corpus below. Posts scoring 70 or above appeared in AI-sourced answer panels at a meaningfully higher rate than posts below that line. It is not an absolute guarantee, but it is the best single-number proxy we have found for "this page is AI-citation-ready."
Which sites are in the corpus?
50 brand blog posts scored on the rubric above, average score 44.4. Click any column header to sort.
| Domain ↕ | Industry ↕ | Score ↕ | Grade ↕ | Date ↕ |
|---|---|---|---|---|
| bankrate.com | Financial Comparison | 80 | B | 2026-05-06 |
| investopedia.com | Finance Education | 71 | C | 2026-05-06 |
| nerdwallet.com | Personal Finance | 70 | C | 2026-05-06 |
| fool.com | Investing Media | 69 | D | 2026-05-06 |
| hootsuite.com | SMB SaaS | 67 | D | 2026-05-06 |
| cloudflare.com | Infrastructure/CDN | 65 | D | 2026-05-06 |
| semrush.com | SEO/MarTech | 59 | F | 2026-05-06 |
| mckinsey.com | Management Consulting | 59 | F | 2026-05-06 |
| sofi.com | Digital Banking | 58 | F | 2026-05-06 |
| buffer.com | SMB SaaS | 57 | F | 2026-05-06 |
| google.com | Marketing Insights | 57 | F | 2026-05-06 |
| clickup.com | Project Management | 55 | F | 2026-05-06 |
| squareup.com | Payments/Fintech | 55 | F | 2026-05-06 |
| onepeloton.com | Fitness Tech | 55 | F | 2026-05-06 |
| datadoghq.com | Observability/DevOps | 52 | F | 2026-05-06 |
| robinhood.com | Investing Platform | 52 | F | 2026-05-06 |
| nike.com | Consumer Brand | 52 | F | 2026-05-06 |
| ahrefs.com | SEO/MarTech | 50 | F | 2026-05-06 |
| figma.com | Design SaaS | 50 | F | 2026-05-06 |
| monday.com | Project Management | 50 | F | 2026-05-06 |
| webflow.com | Website Builder | 50 | F | 2026-05-06 |
| hubspot.com | MarTech | 48 | F | 2026-05-06 |
| zendesk.com | Customer Service SaaS | 48 | F | 2026-05-06 |
| brex.com | Corporate Cards/Fintech | 48 | F | 2026-05-06 |
| stripe.com | Fintech | 47 | F | 2026-05-06 |
| twilio.com | Developer Platform | 47 | F | 2026-05-06 |
| wealthsimple.com | Investing Platform | 47 | F | 2026-05-06 |
| shopify.com | E-commerce Platform | 46 | F | 2026-05-06 |
| duolingo.com | Education Tech | 46 | F | 2026-05-06 |
| zapier.com | SMB SaaS | 42 | F | 2026-05-06 |
| airbnb.com | Travel/Hospitality | 42 | F | 2026-05-06 |
| paypal.com | Digital Payments | 41 | F | 2026-05-06 |
| asana.com | Project Management | 40 | F | 2026-05-06 |
| spotify.com | Music/Streaming | 40 | F | 2026-05-06 |
| canva.com | Design SaaS | 39 | F | 2026-05-06 |
| linear.app | Dev SaaS | 37 | F | 2026-05-06 |
| atlassian.com | Enterprise SaaS | 36 | F | 2026-05-06 |
| miro.com | Collab SaaS | 33 | F | 2026-05-06 |
| salesforce.com | Enterprise | 33 | F | 2026-05-06 |
| intercom.com | B2B SaaS | 30 | F | 2026-05-06 |
| hbr.org | Business Media | 30 | F | 2026-05-06 |
| patagonia.com | Outdoor/Sustainability | 29 | F | 2026-05-06 |
| marketingprofs.com | B2B Media | 27 | F | 2026-05-06 |
| notion.com | PLG SaaS | 25 | F | 2026-05-06 |
| mailchimp.com | SMB SaaS | 22 | F | 2026-05-06 |
| gong.io | B2B SaaS | 21 | F | 2026-05-06 |
| klaviyo.com | MarTech | 14 | F | 2026-05-06 |
| drift.com | B2B SaaS | 10 | F | 2026-05-06 |
| glossier.com | Beauty DTC | 10 | F | 2026-05-06 |
| contentmarketinginstitute.com | B2B Media | 8 | F | 2026-05-06 |
Rubric version: locked-brackets-v2, explicit sub-score brackets, zero interpretive slack. Scored 2026-05-06.