nickel.eu — AI Site Grade

Analysis

Nickel.eu — AI-Visibility Audit

The site's llms.txt returns 403 Forbidden — an active block, not a missing file — while every major AI crawler (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, Bytespider, Applebot-Extended) gets a full 200 with identical content to a browser. The cold LLM knowledge about Nickel is stale by at least two years: it reports "2 million clients" and "pre-2023 data," while the site itself now claims 4+ million accounts opened and has expanded to 5 European countries (France, Spain, Portugal, Belgium, Germany).

Crawler Access

All 11 tested AI bots receive the same 98 KB HTML payload as a browser — no UA-based blocking, no JS shell, no thin content. The site runs on Drupal 10 behind Akamai (2.18.x IPs) with nginx/1.20.1. The robots.txt has a single User-agent: * rule with no AI-bot-specific directives; GPTBot, ClaudeBot, and Google-Extended are not mentioned. The sitemap lives at the non-standard path /sitemap-nickel/sitemap.xml and contains 934 URLs across French, Spanish, Portuguese, German, Dutch, and Belgian subdirectories. The llms.txt endpoint returns a hard 403, which is worse than a 404 — it signals an active nginx-level block that prevents any AI from discovering a content map even if one existed.

Cold-Knowledge Gap

The frontier model queried cold describes Nickel as having "2 million clients" and frames the product as a "no-frills, low-cost current account" targeting "financially excluded individuals." The actual site says 4+ million accounts opened, offers four card tiers (Classic at 25 EUR/year, My Nickel, Chrome at 55 EUR/year, Metal at 105 EUR/year), and positions itself as a mainstream alternative — "le compte pour tous" — not solely a last-resort option. The model also flags "regulatory scrutiny over anti-money laundering compliance" with no date, a claim the site itself does not address anywhere on its public pages. The gap between the model's outdated, risk-tinged prior and the site's growth narrative is significant.

Schema Posture

Only two pages carry structured data. The homepage has a bare Organization schema (name, URL, sameAs links to Facebook/Instagram/TikTok — no logo, no description beyond a short string). The /fr/tarifs-et-services page has a Product @graph with four card offers including price, currency, and availability. Every other key page — the comparison tool (/fr/comparateur), the "how to open an account" guide, the "about us" page, the magazine, and the guide section — has zero JSON-LD. The comparison page, which contains a rich product comparison table with fee structures, is a natural candidate for ProductGroup or Comparison schema but has none. FAQ schema is absent from pages with multiple FAQ sections.

External Signals

The site links to Trustpilot (fr.trustpilot.com/review/nickel.eu) in its footer, but that page blocks automated access with a Cloudflare challenge. No external press or review content was retrievable via web search — the domain has low search-engine footprint for English-language queries. The DNS TXT records reveal deep BNP Paribas infrastructure integration (SPF includes pns._spf.bnpparibas.com) plus Atlassian, Zendesk, and Apple domain verification tokens, confirming a complex enterprise tech stack behind the simple front-end.

Surprising Findings

The /en path returns a 404 with a French error page — there is no English-language homepage despite the site operating in 5 countries. The sitemap contains URLs for de-de, es-es, pt-pt, nl-be, and fr-be subdirectories, but no /en prefix. The magazine section shows articles dated as far ahead as April 2026, suggesting either placeholder publishing dates or a forward-planning content calendar that could confuse temporal reasoning in AI models. The "about us" timeline mentions "2 million clients" as a past milestone (2022) while the homepage now claims "4 million" — the timeline text has not been updated to match the current headline figure.

Findings

1. llms.txt endpoint returns 403 Forbidden High

   The llms.txt endpoint returns a hard 403 Forbidden, actively blocking AI crawlers from discovering a content map. This is worse than a 404 and prevents any AI from using the file even if one existed.

   What to change: Remove the nginx-level block on /llms.txt and serve a valid llms.txt file listing key URLs.

2. Cold LLM knowledge is 2+ years out of date High

   Frontier models describe Nickel as having 2 million clients and a no-frills account for the financially excluded, while the site now claims 4+ million accounts and positions itself as a mainstream option. The model also references unaddressed regulatory scrutiny.

   What to change: Publish an llms.txt file and update structured data to reflect current metrics and positioning.

3. Key pages lack structured data Medium

   Only the homepage and /fr/tarifs-et-services have JSON-LD. The comparison tool, how-to guide, about us, magazine, and guide section have zero structured data, missing opportunities for rich results and AI understanding.

   What to change: Add appropriate JSON-LD (e.g., ProductGroup for comparison, FAQPage for FAQ sections, Article for magazine) to all key pages.

4. English homepage returns 404 Medium

   The /en path returns a 404 with a French error page, despite the site operating in 5 countries. No English-language homepage exists, limiting international AI visibility.

   What to change: Create an English-language homepage at /en or redirect to a localized version.

5. Magazine articles have future publication dates Medium

   The magazine section shows articles dated as far ahead as April 2026, which could confuse temporal reasoning in AI models and reduce credibility.

   What to change: Ensure all published articles have accurate, past or present dates; remove placeholder dates.

6. About us timeline uses outdated client count Medium

   The 'about us' timeline mentions '2 million clients' as a past milestone (2022), while the homepage now claims '4 million'. The inconsistency undermines trust and AI accuracy.

   What to change: Update the timeline text to reflect the current 4+ million figure.

7. Low external search engine footprint Medium

   Web searches for reviews and complaints returned zero results, indicating low visibility in search engines and limited external signals for AI models to reference.

   What to change: Improve SEO and encourage user reviews on platforms that allow indexing.

8. Trustpilot review page blocks automated access Low

   The Trustpilot review page linked in the footer returns a Cloudflare challenge, preventing AI crawlers from accessing user reviews.

   What to change: Consider embedding reviews directly on the site or using a Trustpilot widget that is crawlable.

What's working

• All major AI crawlers receive full HTML content — All 11 tested AI bots receive the same 98 KB HTML payload as a browser, with no UA-based blocking or JS shells.
• Sitemap contains 934 URLs across all locales — The sitemap at /sitemap-nickel/sitemap.xml includes URLs for French, Spanish, Portuguese, German, Dutch, and Belgian subdirectories, aiding discovery.
• Pricing page has Product schema with card offers — The /fr/tarifs-et-services page includes a Product @graph with four card offers including price, currency, and availability.
• Homepage has Organization schema with social links — The homepage includes a bare Organization schema with name, URL, and sameAs links to Facebook, Instagram, and TikTok.
• robots.txt does not block any AI bots — The robots.txt has a single User-agent: * rule with no AI-bot-specific directives, allowing all crawlers full access.