GEO Beat: Data.
What changed, what to do.

The AI search news that matters, with the part the newswires skip: not just what changed, but what to do about it. Written for marketing leads, executives, CMOs and business owners. Every item is verified at the primary source and comes with an operator's take. Three times a week, Monday, Wednesday and Friday, in English and 中文.

中文版 →

// DATA_CATEGORY

Data: the 10 latest briefs

← All GEO Beat briefs

2026-09-02DATA · Medium

Someone else's web page can plant false claims about your brand in AI answers — off-the-shelf safety filters cut that attack's success rate by at most 5.7%.

Someone else's web page can plant false claims about your brand in AI answers — off-the-shelf safety filters cut that attack's success rate by at most 5.7%. A September 2, 2026 preprint, Counter-GEO-Bench (arXiv:2609.02316), ran 247 human-verified queries through three LLMs, each query …

Read the brief →   中文

2026-09-01DATA · High

Only 1.3% of Google Shopping carousel products also appear in AI Mode — so carousel performance does not tell you which price or which seller AI Mode is showing for the same product.

Only 1.3% of Google Shopping carousel products also appear in AI Mode — so carousel performance does not tell you which price or which seller AI Mode is showing for the same product. Productrise tracked 2,000,000+ listings across 100,000+ US and UK queries from August 9–31, 2026: where a …

Read the brief →   中文

2026-08-31DATA · Medium

Profound data study: Claude and Claude Code show divergent web-search and citation behavior

Claude Code searches the web in 13% of responses, Claude in 93% — treat them as separate answer engines. AI-visibility vendor Profound tested 1,724 identical prompts across both products (July 13–23, 2026; 24,135 responses) and found only ~20% brand-mention overlap on the same prompts; a …

Read the brief →   中文

2026-08-18DATA · Medium

Google Analytics logs up to 29% of AI Overview traffic as 'Direct' — and the window worth re-checking is February to June 2026, when AI Overviews peaked at 16% to 17% of organic sessions

Google Analytics logs up to 29% of AI Overview traffic as 'Direct' — and the window worth re-checking is February to June 2026, when AI Overviews peaked at 16% to 17% of organic sessions. A nine-month Search Engine Land study (September 2025 to June 2026; 51,200 citation events across …

Read the brief →   中文

2026-08-11DATA · Medium

Your website has no OpenAI licensing deal and ChatGPT quotes it like Reuters anyway — and what ChatGPT quotes is a page's opening headline plus the next ~150 characters.

Your website has no OpenAI licensing deal and ChatGPT quotes it like Reuters anyway — and what ChatGPT quotes is a page's opening headline plus the next ~150 characters. Two researchers got there separately. French consultancy Resoneo captured 1,249 ChatGPT answers in July 2026 and found …

Read the brief →   中文

2026-08-10DATA · Medium

Your ChatGPT ad buys the slot, not the citation above it — only 3.63% of advertisers are also the source the answer quotes.

SE Ranking ran more than 50,000 commercial-intent prompts across 20 niches and published on August 10, 2026: 25.94% carried a sponsored placement, and the advertiser's exact URL turned up in a citation 0.09% of the time. For business owners and marketing leads: ChatGPT ad spend and …

Read the brief →   中文

2026-07-30DATA · Medium

Your website link now sits under an AI answer on 43–48% of Google searches — you can hold your ranking and still lose the website visit.

Your website link now sits under an AI answer on 43–48% of Google searches. Similarweb and Semrush published charts on July 30, 2026 putting AI Overviews on 43% and 48% of Google queries, five points apart because they measure different crawl samples, and nearly triple the 13% to 20% the …

Read the brief →   中文

2026-07-21DATA · High

More than half of ChatGPT's topic categories have no brand owning the answer — and 90.4% of the brands that do own one keep it the next month.

More than half of ChatGPT's topic categories have no brand owning the answer, and 90.4% of the brands that do own one keep it the next month. Semrush and independent analyst Kevin Indig tested 1,094 U.S. topic categories against five buyer-style ChatGPT prompts from January to June 2026 …

Read the brief →   中文

2026-07-21DATA · Medium

The category word a buyer uses decides whether AI recommends you — the brand's knowledge-graph data is identical either way, so what moves the answer is the category words your content uses

The category word a buyer uses decides whether AI recommends you. A study reported on July 21, 2026 ran 12 athletic apparel brands through ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, 14,140 API runs over seven days, and found lululemon recommended in 90% of answers under …

Read the brief →   中文

2026-07-14DATA · Medium

Buyers engage 1.7x more with niche publishers than with major media — rank your outreach targets by audience affinity, not domain rating alone

A Fractl study published July 14, 2026 (SparkToro data, eight industries) found "hidden gem" niche publishers scored 86–93 on audience affinity at 2,000–10,000 monthly visits, while major publishers often scored 50–65 despite hundreds of thousands. It is vendor-published — Fractl sells …

Read the brief →   中文