The "long articles" myth: where it comes from
Since 2015, the SEO world has been repeating that "long articles rank better". This belief comes from correlation studies (Backlinko, HubSpot) that observed pages in position 1 on Google had an average of 1,890 words. The problem: correlation is not causation.
Long articles often rank better because they cover the topic more in depth, have more internal and external links, and incorporate more secondary keywords. It is the thematic depth that ranks, not the word count.
In 2026, this nuance is critical. Writing 3,000 words of filler to "make it long" is worse than writing 1,200 dense, precise words. Google knows this (Helpful Content Update), LLMs know this (they ignore hollow content), and your readers know this (bounce rate).
"The question 'how many words for my SEO article' is the wrong question. The right question is: 'have I covered all the necessary sub-topics for Google to consider me the best answer?' Sometimes that is 800 words. Sometimes it is 4,000."
2026 data: what the European figures say
Here is the data we have compiled from reliable European sources:
Sistrix (analysis of 50,000 URLs in the top 10, French-speaking market, January 2026):
- Average length of pages in positions 1-3: 2,140 words
- Average length of pages in positions 4-10: 1,680 words
- Sweet spot for informational queries: 1,500 - 2,500 words
- Sweet spot for transactional queries: 800 - 1,500 words
Searchmetrics ("Content Performance Europe 2025" report, DACH + France + Benelux):
- Articles covering more than 80% of identified sub-topics rank an average of 3.2 positions higher, regardless of their length
- "Thematic completeness" is a better predictor of ranking than word count (R-squared of 0.67 vs 0.31)
Length and AI citations: a complex relationship
LLMs do not cite the longest articles — they cite the articles that are most dense in extractable information. But there is an important nuance.
Based on our internal measurements at AI SOS (tracking of 2,400 AI citations over 6 months, French-speaking market), the articles most cited by ChatGPT, Perplexity and Gemini have a specific profile:
- Average length: 2,200 words (range 1,800 - 3,000)
- Number of H2 sections: 5-7
- Presence of tables: 78% of cited articles contain at least one
- Presence of FAQ: 65% of cited articles include a FAQ section
- Informational density: high ratio of facts, figures and definitions vs "bridging" text
The conclusion: LLMs favour articles long enough to be complete, but dense enough to contain no filler. The AI-first format is the best adapted.
Comparison: length by content type
| Content type | Recommended length | Primary objective | AI citability | Example |
|---|---|---|---|---|
| Hub / pillar article | 2,500 - 4,000 words | Exhaustive coverage of a broad topic | Very high | This guide |
| Satellite article | 1,500 - 2,500 words | Depth on a specific sub-topic | High | This article |
| Comparison / vs | 1,500 - 2,000 words | Help choose between 2+ options | Very high (tables) | Semrush vs Ahrefs |
| Case study | 1,000 - 1,500 words | Demonstrate expertise with results | Medium | Client case study |
| News / current events | 500 - 1,000 words | Inform quickly | Low (perishable) | Google update |
| FAQ / glossary | 1,000 - 2,000 words | Answer specific questions | Very high | SEO + AI FAQ |
| Sales page / landing | 800 - 1,500 words | Convert | Low (promotional) | AI SOS offer page |
Method for calibrating the length of each article
Rather than fixing an arbitrary length, use this 3-step method:
Step 1: Analyse the current results
Search for your target keyword on Google. Measure the length of the top 5 results. Identify the sub-topics covered by each. This is your baseline.
Step 2: Identify the required coverage
List ALL relevant sub-topics (Google PAA, Perplexity suggestions, your clients' questions). Your article must cover at least 80% of these sub-topics to have a chance of ranking in the top 3.
Step 3: Write naturally, then edit
Write until you have covered all sub-topics. Then edit mercilessly: remove filler, densify each paragraph, add structures (tables, lists). The final length is a consequence of coverage, not an objective in itself.
"In 12 years of SEO, I have seen 5,000-word articles win and then lose their effectiveness. The 2026 trend is informational density. An extremely dense 1,800-word article outperforms a diluted 4,000-word article, both for Google and LLMs."
FAQ
Does Google penalise articles that are too short?
No, Google does not penalise short articles as such. It penalises content that does not satisfactorily answer the user's query. A 500-word article that perfectly answers a simple question can rank in position 1.
Are articles over 3,000 words too long?
Not necessarily. For complex topics (comprehensive guides, step-by-step tutorials), 3,000-4,000 words can be justified. The test: if you can cut 500 words without losing useful information, the article is too long.
Is the ideal length the same in French and English?
French is approximately 15-20% longer than English for equivalent content (sentences are longer). Adjust your benchmarks accordingly. A 2,000-word article in French is equivalent to approximately 1,700 words in English in terms of informational density.
Should you aim for the same length for all articles in a cluster?
No. The hub is generally longer (2,500-4,000 words) than the satellites (1,500-2,500 words). Length must be adapted to the complexity of the sub-topic being covered. See our cornerstone content guide for pillar pages.
Can AI help determine optimal length?
Yes. Tools like MarketMuse and Frase analyse top 10 content and recommend a target thematic coverage, which translates into a length range. This is more reliable than setting an arbitrary word count.
Are your articles too long? Too short? Poorly calibrated?
We audit the thematic coverage of your content and optimise it for SEO + AI citability.
Audit my content
