Study: 100 French B2B SaaS Sites Analyzed on Geo Score

The French B2B SaaS market remains a blind spot in AI citability research. The few public benchmarks, such as the Ahrefs analysis from March 2026 on 75 000 brands or the Semrush study on 150 000 ChatGPT citations, focus on English-speaking markets. On the French side, Sistrix documented in April 2026 that 58% of Google queries triggered an AI Overview, but without SaaS-specific segmentation. ScoreGeo decided to fill this gap. This article announces the study, details its protocol, sample, and timeline, and opens a call for participation for editors who want to be included or receive their individual score.

Why a study dedicated to French B2B SaaS

No public study currently measures AI citability for French B2B SaaS. Available benchmarks cover either the entire web (Ahrefs, 1 885 pages tested for JSON-LD in March 2026) or the English-speaking market (Semrush, 150 000 ChatGPT citations analyzed). Yet French-speaking B2B SaaS form a homogeneous category, with shared GEO challenges: product documentation, pricing pages, competitor comparisons, reference content.

This gap creates a concrete problem for CMOs and heads of SEO in the sector: no way to benchmark, no way to prioritize GEO work without a reference. The study aims to produce that missing reference, in both data and methodology.

Three axes structure the analysis: technical readability for AI crawlers (GPTBot, ClaudeBot, OAI-SearchBot), semantic content structuring, and off-page authority measured by existing citations in public corpora. The 13 criteria and their weighting are documented on the ScoreGeo methodology page.

Sample: 100 sites, selection criteria

The sample covers 100 B2B SaaS editors based in France, with an active French site and at least one product page in French. The scope excludes pure B2C editors, marketplaces, and agencies. Franco-international SaaS (French HQ, global audience) are included if they maintain a French version of the site.

The distribution targets balance across four segments: horizontal SaaS (productivity, HR, finance), vertical industry SaaS (legaltech, healthtech, foodtech), marketing/sales tools, and technical infrastructure (CI/CD, observability, security). Each segment represents about 25 sites to allow intra-segment statistical comparisons.

Point 1: the final list will be published in full in the report appendix, with individual scores visible. Point 2: an editor can request inclusion via the participation form (priority to profiles completing an under-represented segment). Point 3: no past or current ScoreGeo client is included, ensuring analytical independence.

Methodology: 13 criteria, 100 points, manual audit

Each site is scored on the 13 criteria of the Score GEO framework, for a total of 100 weighted points. The grid combines automated checks (presence of JSON-LD schemas, llms.txt file, robots.txt configuration for GPTBot and ClaudeBot, h-tag structure) and manual audit (answer-first quality of introductions, factual depth, brand mentions in third-party sources). Full methodology is published at scoregeo.ai/methodology.

The audit does not rely solely on scraping: each site is reviewed by a consultant who verifies data consistency and qualifies patterns. This approach limits false positives linked to client-side rendering variations, common on SaaS sites that hydrate content in React or Vue. The typical pattern observed in our offerings shows manual verification changes the final score by 5 to 15 points for about a quarter of sites.

Three criteria historically concentrate most of the gap between well-ranked and poorly-ranked sites: answer-first formatting on key pages, JSON-LD structuring (FAQ, HowTo, Product, Organization), and brand mention density in third-party corpora. Expected dispersion will be widest on these axes.

Timeline and deliverables

Data collection starts in the weeks following this announcement. Phase 1: automated audit of the sample (4 weeks). Phase 2: manual audit and cross-checks (4 weeks). Phase 3: statistical analysis, segmentation, report writing (3 weeks). Phase 4: publication, communication, interactive dashboard release (1 week).

The final report will include: median and distribution of overall scores, breakdown by criterion, segment gaps, the 10 recurring patterns observed, and an appendix with individual scores either named or anonymized per editor choice. A 4-page executive summary will be freely available, the full report will be reserved for newsletter subscribers.

To receive the report early and an invitation to join the sample, subscribe to the ScoreGeo newsletter. Editors wishing to be included can flag their interest in the same form, with priority given to under-represented segments.

What the study will not say

This study does not claim to measure the commercial ROI of AI citations. The link between GEO score and acquisition remains largely undocumented and would require conversion data that editors do not share publicly. The study documents technical and editorial citability, not conversion.

It will not rank SaaS by product performance or customer satisfaction either. The GEO score measures a site's ability to be correctly cited by an LLM in response to a relevant question, which is a necessary but not sufficient condition for visibility. An excellent product with a poor score remains an excellent product, simply less visible in new AI channels.

Finally, the study is limited to a snapshot. AI engine algorithms evolve quickly: GPT-5, Claude 4, Gemini 3 change their citation criteria at every major release. An annual reissue is being considered to track the dynamic.

How this study compares to existing benchmarks

The March 2026 Ahrefs report established that about 28% of web pages carry a valid JSON-LD schema, on a corpus of 1 885 tested pages. The Semrush study on 150 000 ChatGPT citations showed a strong concentration of citations on a few thousand domains. Vercel and MERJ documented 500 million GPTBot fetches across their CDNs, demonstrating the intensity of AI crawling.

None of these works segment by national market or B2B SaaS vertical. The ScoreGeo study therefore positions itself as a vertical and geographic complement to existing horizontal benchmarks. It also draws on the methodology of the foundational GEO paper published in November 2023 by Princeton, Allen Institute, and Georgia Tech, which laid the technical vocabulary of the field.

The chosen format, full audit of each site, is closer to a manual GEO audit than mass scraping. This approach limits sample size but increases interpretive quality, in line with ScoreGeo philosophy: prefer 100 well-analyzed sites to 100 000 poorly measured ones.

How to participate

Three participation modes are available. First mode: receive the report early via the newsletter, with no further step. Second mode: submit your site as a sample candidate, specifying segment, main URL, and a contact. Third mode: contribute by sharing data or qualitative observations on your own GEO strategy, with anonymization on request.

For editors wanting an individual audit outside the study, the ScoreGeo GEO audit offering remains available. To understand how your score compares to the sector median once published, you can cross your audit with the study benchmarks. To go deeper on practical implementation, see the dedicated article on common GEO mistakes and the one on GEO vs SEO distinctions.

All feedback, suggestions for additional criteria, and candidacies can be sent via the newsletter or contact form. The goal is to produce a deliverable useful to the French SaaS ecosystem, with full methodological transparency.

Frequently asked questions