Free tool
Blog GEO Scorer
Paste an article and get a GEO score out of 100, with the specific gaps that stop AI answer engines quoting it.
Rendered HTML gives the fullest picture, because schema markup only exists there. Markdown, MDX and plain text are analysed on the same 100-point scale.
Nothing leaves your browser. The article is analysed on this page and never sent to a server.
What exactly does this tool score?
Eight categories, adding to 100 points. Every check below is one the scorer actually runs against the text you paste. Nothing is inferred, and nothing outside this list affects the number.
| Category | Points | What is measured |
|---|---|---|
| Answer-first opening | 18 | Opening paragraph length, one standalone quotable sentence in the first 150 words, and a TL;DR or summary block near the top. |
| Question headings with answers | 14 | Share of H2 headings phrased as questions, and whether each question heading is followed by a direct answer paragraph of 30 to 70 words. |
| FAQ block | 16 | FAQPage schema (worth 6 of the 16, and only visible in HTML), the number of questions in the FAQ section, and whether the answers stand alone. |
| Tables and lists | 12 | Number of tables, and number of bullet or numbered lists. |
| Paragraph density | 12 | Average sentences per paragraph, and the share of paragraphs running past 3 sentences. |
| Numeric claims | 10 | Count of figures, prices and percentages per 500 words. |
| Outbound citations | 10 | Number of distinct external domains you link out to. |
| Author and date | 8 | Presence of a byline, and a visible publication or update date. |
How honest is the score?
The weightings are our judgement, not a published standard. There is no public algorithm to check them against, so treat the result as a structured checklist rather than a measurement. Where a range is given, such as a 30 to 70 word answer under a question heading, the range is deliberately wide because the precise number is not knowable.
Word count is reported but not scored. A long article is not more quotable than a short one, and the older version of this tool marked anything outside a narrow band as a failure, which was a made-up rule. What matters is whether there is a passage worth lifting, which is what the eight categories above measure.
Each category is normalised to its own maximum, so pasting markdown and pasting HTML give the same ceiling of 100. Where a check cannot run at all, the category is marked not scored and contributes zero. The scorer will not hand out points for content it could not read, which is why a very short paste scores badly rather than scoring well on a technicality.
Why does GEO matter for e-commerce sellers?
When a shopper asks ChatGPT or Perplexity which magnesium supplement to buy, the model picks a handful of sources to quote. Being one of them is a distribution channel in its own right, and it does not follow the same rules as a search ranking. An answer engine needs a passage it can lift and attribute, which is a higher bar than being relevant.
This matters most for sellers who already invest in content, whether that is Shopify store owners or Amazon FBA sellers running comparison guides and buyer content.
How do you improve a low score fastest?
Answer the heading immediately. Rewrite each H2 as the question a reader would type, then put a 30 to 70 word answer directly underneath before any detail. This is usually the single biggest jump in the score.
Add numbers and sources. Two categories, worth 20 points between them, look only for concrete figures and links out to other domains. Replacing vague comparatives with real percentages and citing three independent sources is quick work.
Put a TL;DR at the top. A labelled two or three sentence verdict near the start gives an answer engine the passage it most wants, and it costs you nothing.
Show a byline and a date. Eight points sit on authorship and freshness, and both are template changes rather than writing work.
To track whether the work is paying off, tools like PromptWatch and Otterly.AI monitor whether AI systems cite your brand in their answers.
Frequently asked questions
- What is Generative Engine Optimisation (GEO)?
- Generative Engine Optimisation is the practice of structuring content so that AI answer engines such as ChatGPT, Perplexity and Google AI Overviews can quote it. It overlaps heavily with SEO. The difference is that an answer engine has to lift a passage out of your page and stand behind it, so self-contained passages, sourced claims and clear authorship matter more than they do for a ranking algorithm.
- How does the scorer analyse my blog post?
- It runs eight checks in your browser: the opening, question headings and their answers, the FAQ block, tables and lists, paragraph density, numeric claims, outbound citations, and author and date signals. Each category is normalised to its own maximum, so markdown and HTML share the same 100-point ceiling. Anything the scorer cannot see is marked not scored and left at zero rather than guessed.
- Should I paste HTML or markdown?
- Both work and both are scored out of 100. Rendered HTML is the fuller picture because schema markup only exists there, and FAQPage schema is worth 6 points of the 16 in the FAQ category. If you paste markdown or MDX, that schema check is excluded from the divisor instead of being marked as a failure.
- What GEO score should I aim for?
- Above 70 means the structural work is largely done and the remaining gaps are worth fixing individually. Between 50 and 70 usually means the writing is fine but the article is missing citations, numbers or a direct answer under each heading. Below 50, the article is unlikely to be quoted whatever its quality, because there is no passage an answer engine can safely lift.
- Are these weightings a standard?
- No. They are our reading of what answer engines reward in 2026, based on the patterns we see in the pages that get cited alongside our own. No public algorithm exists to validate them against, so treat the score as a structured checklist rather than a measurement. The word count ranges and thresholds are deliberately broad for the same reason.
- Does this tool store my blog content?
- No. The scorer runs entirely in your browser. Your article is never sent to a server and never stored. You can use it on unpublished or confidential drafts. The PDF report is also generated in the browser.