SEO & AI·6 min read·

Understanding SEO and LLM scores

The two numbers in the editor footer that tell you how well your post will rank on Google and get cited by AI.

Every post in the VeloCMS editor shows two scores in real time: SEO and LLM. Both are graded from 0 to 100. They are independent — a post can score 95 on SEO and 40 on LLM if it is well-optimized for Google but buries its answers. Think of them as two different audiences: the SEO score cares about what search crawlers see, the LLM score cares about what an AI assistant can extract and cite.

Quick start — finding the scores

Open any post in the editor. The scores live in the collapsible SEO panel at the bottom-right of the editor sidebar. Both update automatically as you type, with a short debounce so they're not recalculating on every keystroke. The panel also shows the specific factors that are dragging each score down — you don't have to guess what to fix. Click any factor row and the editor highlights the relevant section of your post.

The SEO score (0-100)

Measures traditional search engine optimization signals — the stuff Google's PageRank and RankBrain have cared about for years. Nine factors are evaluated, each weighted by its established impact on rankings.

  • Title length (50–60 characters is optimal)
  • Meta description length (150–160 characters, must include primary keyword)
  • Heading hierarchy (h1 → h2 → h3, no skipping)
  • Word count (300+ for rankable content)
  • Image alt text coverage
  • Internal and external link density
  • Paragraph length (under 150 words per paragraph)
  • Primary keyword in the first paragraph
  • Readability (Flesch-Kincaid 65–75 for blog content)

A score below 60 usually means one or two structural issues — often a missing keyword in the first paragraph or alt text on images. Scores between 60 and 80 are publication-ready for most blogs. Above 80 and you're genuinely competitive on Google for your target keyword.

The LLM score (0-100)

Measures Answer Engine Optimization — how well your content performs when a language model like ChatGPT, Claude, or Perplexity is trying to extract a direct answer to cite in a response. Seven factors are evaluated.

  • Heading density (more headings = more extractable chunks)
  • Short paragraphs (2–3 sentences each — the sweet spot for AI extraction)
  • List usage (bullet and numbered lists are easy for LLMs to parse)
  • Sentence length (average under 20 words per sentence)
  • Code block presence (signals technical authority)
  • One-idea paragraphs (each paragraph covers exactly one concept)
  • FAQ structure (question-shaped headings trigger featured snippets and AI summaries)

LLM scores below 50 are a sign the content is written for depth-reading humans, not for extraction. Long paragraphs, buried answers, and meandering sentence structure all drag the score down. It's not that the content is bad — it just won't get cited by AI.

How to push both scores above 90

Start with a tight 60-character title that includes your primary keyword. Write an opening paragraph that directly answers the post's core question in one or two sentences — this single change often adds 10+ points to the LLM score alone. Use h2 headings that look like search queries ("How do I...?", "What is...?"). Add three or four internal links to related posts. The AI assistant has an Improve SEO prompt that rewrites your content with these rules applied automatically — useful for a final polish pass before publishing.

Common pitfalls

The biggest scoring trap is writing a long, well-researched article and then burying the answer at the end of the third paragraph. Both scores punish this — SEO because the first paragraph misses the keyword, LLM because the answer isn't upfront. The fix is simple: answer first, elaborate after. Another pitfall is uploading images without alt text. Each untagged image drops your SEO score by 4–7 points depending on the total image count. VeloCMS will flag these in red in the SEO panel. Finally, don't confuse meta description length with excerpt length — the excerpt field is user-facing, the meta description is used in search previews and needs to stay between 150 and 160 characters to avoid truncation.

Related articles

  • How to optimize your blog for AI search (AEO) — the strategic guide behind the LLM score
  • Using the AI writing assistant — use the Improve SEO prompt to automate the fixes
  • Adding images and alt text — keep your image score green

Both scores update client-side — no server round-trip. They run on your post content as-is, including draft text. Publishing does not trigger a re-score; the score at publish time is the score the article ships with.