Analysis
Score a page on five criteria.
Each page gets a score out of 100 and a list of fixes ranked by priority.
- Estimated gain shown for each fix
- Language model score bounded by verifiable rules
Is your content cited by ChatGPT, Perplexity, Gemini or Claude?
Citead scores how citable your published pages are and tells you what to fix to make them easier for AI engines to cite.
The team behind Citead
Citead is built by TW3 Partners
An AI engineering firm in Paris. Companies and schools trust it with their AI projects.
Winner of the VivaTech Challenge Région Sud
Citead, June 2026
Young Innovative Company
TW3 Partners
Osez l’IA catalogue (DGE)
TW3 Partners
France Deeptech member
TW3 Partners
They trust TW3 Partners
Features
What you can do with Citead
Crawl
Find the pages that drag your site down.
Citead reads your sitemap, scores every page and lists the weakest first.
- Score history for every page
- Rescore all pages in one click
Writing
Fix a draft before it goes live.
The editor rescores your text three seconds after you stop typing and highlights the passages to rework.
- Suggested fixes apply in one click
- Direct publishing to WordPress and HTML export for other CMSs
Tracking
Measure what your fixes change.
Citead keeps your site score history and compares it with the competitors you track.
- Competitor scores computed with the same method
- Editorial calendar of upcoming articles
Citation tracking
Know every week whether AI engines cite your site
Citead asks ChatGPT, Gemini, Perplexity and Google AI Overviews your questions and records the sites cited in each answer.
ChatGPT
Gemini
Perplexity
Google AI Overviews
- Your audience's questions
- You save up to 50 prompts per site. Citead checks 10 of them every week on Starter and 25 on Pro, with no credit used.
- Your citation rate, engine by engine
- It is followed week after week and compared with the competitors you track.
- An email when it changes
- After each check, you receive the citations gained and lost.
Citead already works with the tools you use
Google Analytics, Search Console and WordPress connect to your account, and your assistant (Claude Code, Codex or Cursor) works directly on your pages and articles.
Method
Citead publishes its score formula
The table below gives each criterion's weight in the score and shows where a fix earns the most points.
| criterion | sub-measures | share of score |
|---|---|---|
| Semantic richness | entropy, information density, coherence | 52.8% |
| AI citability | quotable sentences, entities, self-contained paragraphs, statistics | 31.4% |
| Freshness | publication and update dates | 8.0% |
| Organization | section diversity, section order, redundancy | 6.0% |
| Authority & citations | source citations | 1.8% |
| Naturalness | wording typical of generated text | not in score |
| Total | 100.0% |
Each sub-measure goes through a square root before weighting: pushing a single measure to the extreme pays less and less. Naturalness is scored separately and does not count in the score.
The free analysis and the public MCP server score by rules, with no language model: the same content always gets the same score. You can test any page, a competitor's included, without creating an account. Inside an account, a language model is added to the same rules, so the score can vary slightly from one analysis to the next.
Research
The score builds on three public studies and on our own paper
The criteria rest on public literature. The scoring algorithm itself is described in a research paper released as an arXiv preprint.
arXiv/September 2025
How to Dominate AI Search
AI engines cite third-party media more than brand sites. The authors conclude that a brand gains authority through coverage in the press and on external sites.
Read the paperPrinceton University & IIT Delhi/ACM SIGKDD 2024
GEO: Generative Engine Optimization
The foundational GEO paper. Researchers showed that adding citations, statistics and quotations increases the likelihood of being cited. Measured on 2023-era generative engines: our own measurements on current engines show weaker effects.
Read the paperarXiv/October 2025
Web Search in The Age of Generative AI
Generative search engines cover a wider range of sources than traditional search, but vary significantly in how much they rely on retrieved content vs. internal knowledge.
Read the paper
Our paper describes and evaluates the Citead score
The paper describes the eleven sub-measures, their weights and the tests that selected them, then compares the score with the citations of ten families of generative engines. Its conclusion: the score works as a quality filter; it does not predict which source will be cited. Method, data and limits are detailed.
- Semantic richness
- 52.8%
- AI citability
- 31.4%
- Freshness
- 8.0%
- Organization
- 6.0%
- Authority & citations
- 1.8%
- Naturalness
- not in score
Real example
Here is the score an article from the Citead blog received.
The overall score and the per-criterion scores are those the Citead engine recorded for this article. The article page shows the same values.
Analyzed article
Le framework d'optimisation de contenu pour maximiser les citations IAThe article is written in French.
Citability score
- Semantic richness92
- AI citability92
- Freshness79
- Content organization85
- Authority and citations93
- Naturalnessnot in the score90
Score calculated by Citead on October 6, 2026
Read the articleHosting and data
Your data is hosted in France.
The server, the database and the models that write and score your articles run at Scaleway, in its Paris region.
- Backup retention
- 30days
- One backup a day, encrypted by the host, is kept in the European Union and then replaced.
- Text stored
- 0
- The free analysis and the public MCP server score the text without an AI model and do not store it.
- Model training
- Art.12.2
- This article of our terms rules out using your content to train a model, by us or by our providers.
Parisfr-par
France hosts the server, the database and the models.
Encrypted backups stay in the European Union.
Outside this map, Brave and Resend process some data in the United States.
How your data travels
From your account to the subprocessors called for specific tasks.
You
Your site and your content
The sites, pages and articles in your account, with their history.
Citead, in Paris
Scaleway infrastructure
- Application
The server and the database are in France, in the fr-par region.
- Models
DeepSeek V4 Flash
Served by Scaleway Generative APIs for analysis, scoring and writing.
Your content trains no model (terms, Art. 12.2).- Backups
One a day, encrypted, kept 30 days in the European Union.
JSON export
You download your sites, pages, articles and history from your account settings.
Account deletion
Your data is erased from the database, then from backups within 30 days.
- AI Act notice
Article generation displays the notice required by Article 50.
Specific tasks
These subprocessors handle one precise task each, some outside the European Union.
- Google GeminiEuropean UnionGemini models for video analysis, image generation and research in the editor.
- DataForSEOKeyword and search result data, citation tracking in AI answer engines.
- BraveUnited StatesWeb search for automatic article writing.
- Jina AIReads public web pages when the internal extractor fails.
- StripeEuropean UnionPayments and subscription management.
- ResendUnited StatesSends transactional emails.
- PostHogEU, FrankfurtAudience measurement, after your consent.
Pricing
Starter covers one site and Pro covers five
Analysing a page stays free. The subscription adds the crawl, tracking and content generation.
Starter
For tracking and improving a single site
29€/month
that is €0.97 a day
Up to 50 articles a month for 1 site
The plan includes 50 credits a month. A written article costs 1 credit and a rewrite 0.5 to 2 credits depending on its scope. Analyzing a page is free.
- Unlimited page analyses
- Score and fixes in real time
- Recommendations per criterion
- Competitor tracking (10 per site)
- AI citation tracking: 10 prompts per site, checked every week
- MCP server and personal API keys
- Email support
Recommended
Pro
For teams running several sites
79€/month
that is €2.63 a day
Up to 200 articles a month for 5 sites
The plan includes 200 credits a month. A written article costs 1 credit and a rewrite 0.5 to 2 credits depending on its scope. Analyzing a page is free.
- Everything in Starter, plus:
- AI citation tracking: 25 prompts per site, checked every week
- Full crawl, score per page
- Editorial calendar
- Priority support
Enterprise
For more than 5 sites or more than 200 articles a month
Custom
- Everything in Pro, plus:
- Dedicated account manager
- Training for your teams
14-day free trial on Starter and Pro, no commitment. Prices before tax.
FAQ
Frequently asked questions about Citead and GEO
Contact
Write to us about your site or the method
Tell us what you are working on. We reply within one business day.
How it works