Yext AI (Scout and Yext Agents)

par Yext

AI Application

Faible confiance

Yext AI is the generative and agentic layer of the Yext digital presence platform, used by multi-location brands (retail, restaurants, hospitality, healthcare, financial services) to manage how their locations, products and professionals appear in search engines, maps, review sites, social networks and AI assistants. Brand facts sit in the Yext Knowledge Graph and are pushed to over 200 publishers (the 10-K names Google, Apple, Bing, Facebook, Yelp, Amazon Alexa, Gemini and OpenAI). Scout, described by the vendor as an AI search visibility agent, measures how the brand and each location appear in Google Search, Maps and AI Overviews and in ChatGPT, Gemini, Claude and Perplexity, benchmarks up to 20 competitors, analyses sentiment and citations, and recommends fixes. Yext Agents (Scout, Knowledge Graph, Listings, Reputation and Social agents) can then draft and publish review responses, social posts and content, correct listing data and sync updates, with an Action Center where staff can review or step in. Related products include Pages, site Search (natural-language answers on the customer website), Reviews and Hearsay Social and Relate for financial advisors. Vendor-reported scale figures (for example '10B signals, 12M+ locations') are not verified. Risk rationale: rated limited. The system does not fall in an EU AI Act Annex III category: it manages public business information and marketing content, not decisions about people's access to jobs, credit, essential services or similar. The obligations that do apply are the Article 50 transparency duties: website Search and chat experiences interact directly with consumers, and agents generate review replies, social posts and page text published in the brand's name, so deployers should disclose AI interaction and keep humans approving public outputs. Personal data is mostly ordinary (reviewer names and review text, end-user search queries, customer staff accounts, professionals' names and photos). Assurance is reasonably strong for a SaaS vendor: SOC 2, HIPAA, PCI DSS, GDPR and CPRA are listed in a SafeBase Trust Center (vendor statements, reports behind access), EU-U.S. DPF certification, EU hosting options in Frankfurt, Amsterdam, Warsaw and Mons per the 10-K, and public incident notices. Open points: the LLM providers used today, AI training-data terms and the sub-processor list are not public; agents can publish to third-party sites at scale, so an error propagates quickly; two 2026 incidents touched peripheral systems (Klue-linked CRM access, placesscout.com). The rating would rise to medium if a deployment uses Hearsay Relate or Search in healthcare or financial services where consumer queries or messages carry special-category or regulated data, or if agents publish without human approval; it would stay limited where outputs are reviewed and only public business data is processed.