Latest AI search research

Brand visibility in Google AI Overviews is most strongly correlated with the frequency of positive branded mentions across authoritative third-party sources.

Planted text comprising less than 4% of page content still successfully surfaced in AI-generated reports in 30-53% of sessions.

A randomized field experiment found no measurable difference in bounce rates, return to search, or time on site between clicks with AI Overviews and those without, contradicting Google's claim that lost clicks were low-value.

Google's official position states that AI model training on publicly available web data qualifies as transformative fair use under U.S. copyright law.

Over 50% of active ChatGPT users (Free, Go, Plus, Pro plans) predominantly use non-English languages, with Spanish, Portuguese, and Arabic being most common.

ChatGPT directs 28.8% of its referral traffic to internal search pages, with approximately 25% of all AI-referred traffic landing on internal search across verticals.

36% of fintech websites (274 sample) are partially invisible to AI crawlers due to JavaScript rendering requirements; 17% deliver zero content without JS execution.

Click quality metrics (back-button rate, 10-second abandonment, time on site) showed no statistically significant difference between conditions where AI Overviews were present versus absent.

Google's official Computer Use safety documentation identifies unique security risks from untrusted content and model execution errors.

Original data is the strongest correlate of information gain score; pages with 15+ unique figures score 62.1 vs. 40.2 for pages with ≤1 unique figure.

Google's quality principles for websites are not expected to change significantly for agentic AI browsers; sites optimized for human users will generally work for AI agents.

Google's official guidance recommends meta descriptions for critical pages but states they are not required for all pages.

Reid states that personalized search signals can surface niche publishers and specialist reviewers that generic one-size-fits-all results would not display.

An iPullRank controlled experiment with three account variations running identical prompts with different personalization levels produced measurably different AI Mode responses, including personalized name-specific outputs.

In template-driven store layouts, browsers can incorrectly identify the LCP element, causing Core Web Vitals optimization efforts to target the wrong element.

Google launched a visual treatment in AI Mode that displays prominent recipe links at the top of responses for relevant recipe queries with metadata including creator name, ratings, and ingredient counts.

Automated bots comprised 57.2% of HTTP requests to HTML content during May 30 to June 5, 2026, crossing the 50% threshold before the previously forecasted 2027 date.

AI training crawler traffic grew from approximately 20% in spring 2025 to become the majority of crawler requests on Cloudflare's network; daily AI agent requests increased by over 1,700% year-over-year.

Median agent run took 32 seconds and $0.24 with three tool calls; fastest to slowest runs showed 2.2x time difference and 4.2x cost difference, primarily due to web-search calls.

Google's June 2026 spam update includes enforcement against tactics that manipulate generative AI responses, including buying or altering citations for AI answers.

Google officially released the June 2026 spam update on June 24, 2026, which completed rollout by June 26, taking approximately two days.

Approximately 25% of U.S. adults use AI chatbots on a daily basis.

Branded desktop searches gained 1.99-5.78 percentage points across top-10 positions; unbranded mobile #1 position dropped 3.07 points.

Brands cited in AI Overviews experienced 120% higher organic clicks per impression compared to uncited competitors on the same SERP; uncited brands saw 67% CTR decline year-over-year.

Google announced the June 2026 spam update on June 24, 2026, applying globally to all languages with a rollout timeline of a few days.

Users who received a ChatGPT brand recommendation were 2.5x more likely to visit that brand's website within 7 days compared to competitor websites.

User-generated platforms account for 17-23% of all retrieved URLs across tested systems, with Reddit comprising 54-71% of user-generated content citations.

97% of llms.txt files across 137,000 domains received zero requests from AI citation bots, with citation-generating bots comprising only 1% of total fetches.

Converting HTML to text is trivial and does not present a technical barrier for crawlers, contrary to claims that markdown simplifies content extraction for AI agents.

Users arriving via ChatGPT recommendations engaged 85% deeper (12 vs 6.5 pages) and stayed 110% longer (11.8 vs 5.6 minutes) than users from other sources.

80% of AI crawling activity is associated with model training versus search or user-driven crawls

In 224 of 323 citations (69%), Google cited a brand's self-promotional listicle but excluded that brand from AI Overview recommendations, instead recommending competitors.

Google Cloud published reference implementations including an enrichment agent for producing OKF documents from BigQuery datasets with citations and schema documentation.

Google states that converting HTML to text for LLM consumption is trivial, contradicting the premise that markdown simplification is necessary for AI crawling.

Reddit is cited with authoritative medical sources 36% of the time in ChatGPT but only 6% of the time in Google AI Overviews, indicating AI engines assign different contextual roles to the same source.

92% of ChatGPT citations are not cited by Claude for the same query, indicating significant algorithmic divergence between AI systems.

Meta launched AI Mode in Facebook Search that generates answers powered by Meta AI and uses public social content from Groups, Reels, and other Meta apps as sources.

Fable 5 achieved the highest score among frontier models on Cognition's FrontierCode evaluation for difficult coding tasks.

Expert survey of ~50 local SEO professionals found that complete and current business information is among the most important signals for local search rankings.

Google confirmed that hyphenated domain names carry no negative SEO signal, and domain names can contain up to 61 hyphens.

300 marketing executives from large enterprises were surveyed regarding AI search adoption and organizational response strategies.

Social search adoption among Americans increased from 30% in 2020 to 35%, representing a five-percentage-point shift in information discovery behavior.

Reddit was cited alongside medical authority sites 36% of the time in ChatGPT but only 6% in Google AI Overviews, representing a 6x difference in how the same source is contextually positioned.

Only 2.2% of citations remain stable across three identical ChatGPT prompt runs from a dataset of 815,000 prompt-page pairs.

SISTRIX visibility analysis across US and UK markets (May 26 - June 2) showed sites matching query intent and result type gained visibility while mismatched sources lost ground.

Google rolled out a new Generative AI report section in Search Console showing impressions data for content appearing in AI Overviews and AI features, with filtering capabilities by pages, countries, devices, and dates.

Reddit's top 3 SERP share increased to 10.24% after May 2026 core update, up from 8.56% after March and 9.19% after December, across 100,000 tracked keywords.

Brands with presence on four or more platforms showed a 2.8x increase in likelihood of appearing in ChatGPT responses compared to those with fewer platform mentions.

Google announced that users can follow creator profiles to increase visibility of that creator's content in the Discover feed.

Google officially stated that third-party SEO tools are not endorsed by Google and lack access to Google's internal ranking data.

61% of AI users cite speed and ease as primary reasons for using AI tools.

AI Overviews compress time-on-SERP behavior across search intents: without AIOs, navigational vs. local searches show 20-point variance (12% vs. 32%), but with AIOs all five intent types cluster between 41.9-48.5% at 21 seconds.

Traditional search outpaces AI tool growth in absolute volume despite faster growth rates for AI tools.

Study used three AI-detection tools to classify content across a 6-year publication span from January 2020 to March 2026

AI-referred traffic converted 42% better than non-AI traffic in March 2026, representing a sign flip from worst-performing to best-performing channel in U.S. retail within 12 months.

Google CEO Sundar Pichai acknowledged that AI Overviews require further tuning and refinement in how they answer certain queries, as some responses contain excessive opinion.

Google rolled out new AI performance insights in Merchant Center to track product performance across AI-powered shopping experiences, including share of voice benchmarking, funnel performance metrics, conversational query insights, and product attribute completeness tracking.

Adding JSON-LD schema to 1,885 pages resulted in a 4.6% decline in AI Overview citations relative to control pages, with AI Mode and ChatGPT showing statistically insignificant changes of +2.4% and +2.2% respectively.