AI-Generated EMDs Outranked a French Government Portal in LLM Citations

Spam sites slip into ChatGPT responses because the platform appears to lack mature anti-spam detection models like Google's.

Victor Da Silva observes that 100% AI-generated exact match domains (EMDs) outrank the French government portal Francenum.gouv inside LLM citation sources for "best electronic invoicing software," a query carrying €20 to €30 CPCs.

Operators deploy coded comparison EMDs — such as an "electronic invoice comparison" domain — to manufacture web consensus around high-value commercial terms.

In Da Silva's view, this window stays open only while ChatGPT and Claude focus on building their business models; once established, the platforms will construct their own clean web indexes and deploy anti-spam filters to purge these sites.

Separately, aggregating LLM citations across a topic-categorized scope of 30 prompts exposes a tight concentration of authority.

While GEO tools sometimes surface 100, 200, or 300 sources per pull — and the 30-prompt batch might mean analyzing roughly 400 to 500 sources — 10 to 15 core sources stand out each time.

Auditing each core source divides targets into competitors and third parties.

When a competitor like Penny Lane holds a core source slot with a dedicated "electronic invoicing software" page where getting a mention is not possible, the counter-move is creating or optimizing a matching page on the brand's own domain.

When the core source is a non-competitor comparison site, media outlet, or external company, brands secure placements through reciprocal link-and-mention swaps, direct outreach, or link marketplaces.

Publishers holding those top slots price this bottleneck accordingly: in the electronic invoicing market, sites created just 1 year ago — rather than legacy media outlets like Capital or Ouest-France — command €2,000, €3,000, to €4,000 for a single link or brand mention.