Skip to content

ANALYSIS · WORK

AI and job postings in France: what the Coface and France Travail studies actually measured

On 17 November 2026, Coface and France Travail published two studies on AI and hiring in France. One counts job postings that never appeared, the other records what employers expect within five years. Neither proves that AI caused the gap.

Lire cet article en français

Reception hall of an office building, with a desk under the Coface logo and a revolving door in the foreground
The reception of Coface’s head office in Bois-Colombes, near Paris, photographed in June 2015. The trade-credit insurer published the AI-LERT indicator with the OEM on 17 November 2026. Photo: Swaf75, CC BY-SA 4.0, via Wikimedia Commons, cropped (same licence)

In brief

  • According to Coface and the OEM, 10 extra points of AI exposure went with 12.4% fewer job postings in the third quarter of 2026 than in less-exposed occupations of the same industry, relative to before ChatGPT.
  • Coface’s own data file shows that gap between −11.8% and −12.7% since late 2025: a plateau, not an acceleration.
  • In France Travail’s survey of just over 16,000 establishments, 6% of employers expect AI to take over many tasks of their main occupation within five years; 59% expect little or no effect.

Two French studies landed on 17 November 2026. The headlines said AI was eating jobs. The documents are more careful than that. And one of the most quoted figures appears in neither.

Autumn 2022: the baseline, just before ChatGPT

Coface, the French trade-credit insurer, built its AI-LERT indicator with the Observatoire des emplois menacés et émergents (OEM), a research group on threatened and emerging jobs. It tracks job postings filed with France Travail, France’s public employment service, against their level in the third quarter of 2022. Before that date, according to Coface, there was no statistically significant link between an occupation’s AI exposure and its postings.

2023 to 2025: from the back office to sales

The gap opened in the third quarter of 2023, according to Coface’s press release (in French). The data file published alongside it traces the slope: −6.3% in the third quarter of 2024, −10.5% a year later, −12.2% at the end of 2025. Coface’s English-language report places the first hits on secretarial, administrative and accounting functions in 2024, with sales-related occupations weighing more from 2025 onwards.

That second shift is the one a marketing team in France should read closely. The report singles out telesales and telemarketing: in the third quarter of 2026, postings were 5% below sector-adjusted expectations in office and business support services, where those roles are central, but 39% below in the rest of the economy. Software developers show the same split, −22% in IT and engineering against −40% elsewhere. In other words, the cuts land hardest where the job sits on the edge of the business, not at its core.

Temporary employment agencies amplified the decline, Coface adds: for every 10 extra points of exposure, their postings show an additional 15% drop in relative posting odds compared with the rest of the market, according to the same report.

17 November 2026: two studies, and a headline that ran ahead

For the third quarter of 2026, Coface reports that 10 extra percentage points of AI exposure went with 12.4% fewer postings than in less-exposed occupations of the same industry, relative to pre-ChatGPT levels. Sales, accounting, HR, IT, consulting, legal and R&D sit 20% to 35% below expected levels in the key figures of the press release; its body says 20% to 30%. Germany (−10.8%) and Sweden (−10.2%) show a similar pattern. AFP, in a dispatch carried by France 24 (in French), swapped the two countries and wrote of 25,000 fewer ads in one quarter, a figure found neither on Coface’s page nor in its release.

Now the objection. The data file does not show a collapse that keeps accelerating: −12.7% in the first quarter of 2026, −11.8% in the second, −12.4% in the third, inside a confidence band of roughly −9% to −16%. Four quarters on a plateau. Coface itself writes that its results do not isolate the contribution of AI.

The same day, France Travail’s labour market observatory published a survey (in French) of just over 16,000 establishments. Only 6% of employers expect AI to carry out many tasks of their main occupation within five years, 35% expect time savings and 59% little or no effect. For communication, information, arts and entertainment, the split is 12%, 46% and 42%. Among employers already using AI, 70% list generative AI skills as something to develop, according to its infographic (in French). For how many French firms deploy AI at all, see our reading of the official data on AI use at work, and more of this beat in our Society section.

By 2031: the horizon employers were asked about

France Travail’s question looked five years ahead. In its survey, 74% of employers of communication and information professionals expect AI to change the tasks of the job by then, and 58% of those employing sales representatives. These are expectations, not measurements. For a marketing team that leaves a telemarketing post unfilled and automates prospecting instead, the consent rules do not change: see our guide to AI sales prospecting under CNIL and GDPR rules. The same goes for generated campaign visuals, covered in our guide to labelling AI-generated ads.

Still undated

The documents read on 21 November 2026 give no date for the next AI-LERT update. No figure for marketing roles as such. And no estimate of how much of the gap AI itself explains.

Frequently asked questions

Did AI wipe out 25,000 job ads in France?

The figure of 25,000 fewer ads in one quarter comes from an AFP dispatch of 17 November 2026; it is not in Coface’s French web page or press release. Coface estimates more than 100,000 missing postings a year in the most exposed families and says it cannot isolate AI’s contribution.

Are only entry-level jobs affected?

Not according to Coface. By the second quarter of 2026, the gap was −14.0% for roles requiring under a year of experience, −19.4% for one to two years and −16.0% beyond four years. Coface says the differences are not statistically significant.

Do Germany and Sweden show the same pattern?

Yes, using a comparable method: Coface finds −10.8% in Germany and −10.2% in Sweden for 10 extra points of AI exposure in the third quarter of 2026. The AFP dispatch swapped the two countries.

What is France Travail?

France’s public employment service, which registers job seekers and collects job offers from employers. Its labour market observatory ran the employer survey, and Coface built its indicator from the postings filed with it.

Share this article

Primary sources

  1. AI-LERT: Tracking the Jobs that Never Were (Coface Economic Publications, Focus) — Coface, published 17 November 2026 · accessed on
  2. IA et emploi : quatre ans après le lancement de ChatGPT, les recrutements ralentissent davantage dans les métiers les plus exposés — Coface, published 17 November 2026 · accessed on
  3. Communiqué de presse : IA et emploi, quatre ans après le lancement de ChatGPT, les offres d'emploi des métiers les plus exposés décrochent — Coface, published 17 November 2026 · accessed on
  4. AI-LERT : écart d'évolution des offres d'emploi (données du graphique, .xlsx) — Coface et OEM, published 17 November 2026 · accessed on
  5. L'intelligence artificielle et les métiers à 5 ans : la vision des employeurs (Éclairages et synthèses n° 91) — France Travail, published 17 November 2026 · accessed on
  6. L'impact de l'IA à 5 ans sur les métiers : la vision des employeurs (infographie) — France Travail, published 17 November 2026 · accessed on
  7. IA : dans les métiers les plus exposés, une importante baisse des offres d'emploi, selon une étude — AFP (via France 24), published 17 November 2026 · accessed on

An article by the Rédaction Société & Droit, published under the editorial direction of Bastien Hugon. Editorial method and use of AI

Also read