Latest AI search research

Search results layout has changed significantly over seven to eight years, with AI overviews and commerce links now dominating above organic results.

Google officially launched SynthID verification in Search with planned Chrome rollout in coming weeks.

Google reported AI Mode has surpassed 1 billion monthly active users with query volume doubling quarterly since launch and reaching an all-time high in the most recent quarter.

Google documentation states that absence of llms.txt may increase agent crawl time to understand site structure and primary content.

A controlled experiment tracking AI bot visits to webpages with LLMs.txt files over a 3-month period found zero AI bot visits to the test pages.

Web Bot Auth is in experimental phase with limited rollout; not all Google user agents currently implement the protocol.

AI search engines collectively drive approximately 1% of publisher traffic.

Citation overlap of sources varies from 16-59% across AI engine pairs, but brand overlap remains consistently narrow at 35-55%, indicating AI engines cite different sources but recommend the same brands.

GPT-5.5 achieved an 84.9% win-or-tie rate against human professionals across 44 tested occupations in the GDPval benchmark.

Year-over-year LLM referral traffic increased 72.7%, with ChatGPT referrals growing 66.7% and Claude referrals growing 23x.

Real-time answer fetching accounted for 56.9% of crawler activity (primarily ChatGPT), training 28.8%, and discovery/indexing 14.3%.

48% of AI citations originate from community platforms (Reddit, YouTube), and 85% of brand mentions in AI answers come from third-party pages rather than brand-owned domains.

Consistently pairing a brand name with a descriptive expertise label ('unoccupied home insurance experts') across BBC News and The Times coverage produced a 194% increase in GEO traffic and 382% rise in ChatGPT referral visits — demonstrating the context wrapping technique.

AI brand recommendations are highly inconsistent: fewer than 1-in-100 chance ChatGPT or Google AI returns the same brand list across two runs of the same prompt, based on 2,961 prompt tests with 600 volunteers across 12 categories.

YouTube receives 200x more AI citations than any competing video platform (Vimeo, TikTok, and Dailymotion each below 0.1%), accounting for 20% average citation share and 29.5% of all Google AI Overview citations — ranking as the #1 cited domain in AI Overviews.

Perplexity failed to read JavaScript-rendered pages, returning Access Denied errors, while server-side rendered comparison pages were successfully parsed — demonstrating consistent AI crawler inability to process client-side rendered content.

Anthropic's ClaudeBot crawls web content for potential use in AI model training; site owners can block it via robots.txt using the 'ClaudeBot' user agent string, with IP verification available at Anthropic's published endpoint.

OpenAI operates two separate crawlers: GPTBot for training data and OAI-SearchBot for ChatGPT search results — each requires independent robots.txt configuration; allowing one does not enable the other.

Core Web Vitals are a Google ranking signal but act as a floor rather than a ceiling: Google will surface sub-par page experience content if it is the most relevant result available.

Bing-indexed content is included in Bing Chat (Copilot) answers by default; sites must use the NOARCHIVE meta tag to opt out — confirming that Bing indexing status directly controls inclusion in Copilot AI answers.

ChatGPT Search for Enterprise and Edu workspaces uses Bing as its sole third-party search provider — confirming Bing indexing is a prerequisite for ChatGPT visibility in enterprise contexts.

Microsoft 365 Copilot uses the Bing search service for web grounding; URLs not indexed by Bing cannot be used as knowledge sources — making Bing indexing a hard prerequisite for Copilot visibility.

Google's core ranking systems require LCP under 2.5 seconds, INP under 200ms, and CLS under 0.1 as Core Web Vitals thresholds; achieving these is described as necessary for success in Google Search including AI features.

Microsoft reported a 357% year-over-year increase in AI referrals to top websites in June 2025, totaling 1.13 billion visits.

86% of AI citations across ChatGPT, Gemini, and Perplexity come from brand-managed or brand-influenced sources: 44% from brand websites, 42% from listings/directories, and 8% from reviews/social, based on 6.8 million citations from 1.6 million queries.

Reddit accounts for 6.6% of Perplexity citations versus 1.8% for ChatGPT across 680 million citations analyzed — platform-specific citation preferences differ substantially, requiring off-site presence across multiple platform types for cross-platform AI visibility.

Google officially recommends including author name and a unique author URL as Article structured data properties, with author.url specified as 'a link to a web page that uniquely identifies the author.'

Brands cited in Google AI Overviews receive 120% more organic clicks per impression than non-cited brands on the same AIO-present SERP (2.07% vs 0.94% organic CTR for informational queries), based on 5.47 million queries and 2.43 billion impressions across 53 brands.

Government (.gov) websites represent 6% of sources linked in Google AI Overviews versus 2% in standard search results — a 3x overrepresentation — based on analysis of 68,879 real searches by 900 US adults, indicating AI Overviews heavily favor authoritative government domains.

Perplexity officially confirms PerplexityBot surfaces and links websites in Perplexity search results and is not used for AI foundation model training — distinct from training crawlers.

Approximately 50% of Perplexity citations link to 2025 content and 80% to 2023–2025 content, based on analysis of 5,000+ URLs, making recency the dominant Perplexity citation factor.

Expert authors with credentials and biographies achieved a 72% AI citation rate vs 25% for unattributed content — a 2.4x lift — across 1,200+ pages and 3,600+ queries on four AI platforms over 90 days.

Ahrefs analysis of 75,000 brands found branded web mentions (0.664 Spearman correlation) are 3x stronger predictors of AI Overview citation than backlinks (0.218) — the single strongest signal for AI search visibility is how often your brand is mentioned across the web.

Person schema appears in 58.9% of AI-cited pages (70.4% for ChatGPT specifically), making it the most prevalent schema type in citations, likely because it signals author authority to AI systems.

HTTP Archive analysis of 12 million websites shows AI crawler blocking grew from zero to 125,000 sites within one month of GPTBot's documentation — and continued accelerating, with higher-traffic sites most likely to implement AI bot restrictions.

ClaudeBot crawls 20,583 pages for every single referral it sends back — compared to Google's 5:1 ratio — meaning AI crawlers consume content heavily but return almost no traffic, making citation-based visibility the only meaningful outcome of AI search indexing.

SE Ranking's XGBoost model trained on 300,000 domains found that removing llms.txt as a variable improved prediction accuracy — indicating llms.txt has no measurable positive effect on AI citation frequency as of late 2025.

Comprehensive content (10,000+ words, Flesch readability score ~55) received 62x more AI citations than thin equivalents (187 vs 3 citations) — indicating content depth and readable comprehensiveness are major citation drivers.

In a 90-day experiment with 62,100+ AI bot visits, llms.txt received only 84 visits (0.1% of crawler traffic) and performed 3x worse than average content pages — providing no measurable crawl priority benefit.

Semrush analysis of 100M+ AI citations found only 11% of domains are cited by both ChatGPT and Perplexity — meaning optimising for one AI platform does not automatically confer visibility on another, requiring platform-specific strategies.

AI crawler traffic grew 96% in one year (May 2024–May 2025), with GPTBot's share of crawler traffic rising from 5% to 30%, based on Cloudflare analysis.

Analysis of 1.2M ChatGPT responses found 44.2% of citations are drawn from the first 30% of page content — meaning leading with answers and key facts is critical for AI citation, not burying them mid-article.

Google reports that AI Overviews correlate with meaningfully longer and more natural language queries from users.

In an analysis of over 26,000 ChatGPT source URLs, 'best X' listicle pages accounted for nearly 44% of all content types cited by ChatGPT, including self-promotional rankings.

Gemini and ChatGPT exhibit near-opposite citation and mention behaviours: Gemini mentions brands in 83.7% of appearances vs. ChatGPT's 20.7%, while ChatGPT cites 87.0% vs. Gemini's 21.4%, meaning platform-specific optimisation strategies are required.

ChatGPT citation rate drops from 58% at retrieval position 0 to 14% at position 10, making retrieval rank the strongest predictor of citation.

OpenAI, Anthropic (ClaudeBot), Meta, ByteDance, and Perplexity crawlers do not render JavaScript. Google (Gemini) and Applebot are the only AI crawlers with full JavaScript rendering. ChatGPT hits 34.82% 404s per crawl session; Claude 34.16%. Prioritize SSR/ISR/SSG for main content, meta information, and navigation. Client-side rendering only for enhancement features.

OpenAI officially recommends WAI-ARIA roles, labels, and states on interactive elements as a prerequisite for ChatGPT Atlas to correctly interpret and interact with web pages.

ChatGPT-User crawler requests increased 2,825% year-over-year in 2025, with AI retrieval crawling growing more than 15x across the year as measured by Cloudflare's network.

Google has patented a persistent background search system (US12536233B1) that stores unanswered user queries and proactively delivers results across future sessions and devices — including unrelated assistant conversations. The system eliminates the need for users to repeat searches by tracking information needs over time. This represents a fundamental shift toward agentic, proactive search and has GEO implications: content addressing specific unanswered questions is more likely to be surfaced in these autonomous delivery cycles, making direct-answer-format pages and comprehensive FAQ content more valuable.

Google AI Mode in Chrome now opens cited pages side-by-side with the AI conversation, allowing users to compare details and ask follow-up questions without losing context. This change increases the click-through value of AI Mode citations: pages cited in AI Mode are now directly visited more often because the friction of switching tabs is removed. For GEO, this reinforces the importance of being cited in AI Mode — citations now deliver direct page visits, not just AI summary impressions. Pages that are crawlable, mobile-responsive, and load quickly will benefit most from this change.

AI-sourced visitors now convert 42% better than other digital channels — a dramatic reversal from a prior period when AI traffic converted 38% worse than average. Revenue per visit from AI referrals is 37% higher, and AI traffic has grown 393% year-over-year across industries. This data establishes the business case for GEO investment: AI search does not merely drive volume, it drives higher-quality visitors who are further into the buying journey when they arrive. Brands optimising for AI visibility are capturing visitors with higher purchase intent.

JSON-LD structured data — Google's recommended schema implementation format — is invisible to all 6 AI crawlers tested (0/6 readability score). Meta descriptions and Open Graph tags score equally low: 0/6. Only the <title> tag achieves reliable readability across AI crawlers (5/6). This directly challenges the assumption that implementing schema markup improves AI citation rates through metadata signals. The practical implication: schema markup may benefit traditional search crawlers and humans but does not influence how current AI crawlers ingest page metadata. Content signal, not metadata signal, drives AI discoverability.

Aligning page titles and URLs to ChatGPT's narrower sub-queries correlated more strongly with citation selection than broad keyword matching alone.

The original large-scale empirical confirmation that ChatGPT's citation engine draws primarily from Bing, not Google. 87% of SearchGPT citations match Bing's top organic results; Google achieves only a 56% match at a median citation rank of 17 — meaning pages ranking deep in Google still get cited if they rank well in Bing. This finding, corroborated by subsequent studies, establishes Bing SEO as a distinct and necessary investment for any brand seeking ChatGPT visibility. A Google-only SEO strategy leaves 31% of the citation surface unaddressed.

Bing search rankings predict ChatGPT citation rates more reliably than Google rankings. A hotel with dominant Bing presence was cited 19% of the time; a comparable hotel with stronger Google rankings but weaker Bing presence was cited only 1.5% of the time — a 12x difference driven entirely by which search engine ChatGPT draws from. The study corroborates earlier data showing 87% of SearchGPT citations align with Bing's top results. For ChatGPT visibility, Bing SEO is not optional — it is the primary ranking signal.

LinkedIn is the second most cited domain across all three major AI search platforms — ChatGPT Search, Google AI Mode, and Perplexity — appearing in 11% of AI responses on average. This establishes LinkedIn as a primary AI citation surface, not a secondary channel. AI responses show semantic similarity scores of 0.57–0.60 with original LinkedIn content, indicating substantial direct quotation and paraphrase. For brands and professionals, LinkedIn content is not just a social signal — it is crawled, indexed, and cited by AI search at near-domain-level frequency, making it a first-class GEO publishing surface.

In AI Mode, 64% of users make high-stakes purchase decisions without clicking a single link — a complete inversion of traditional search behavior where 89% click through. This means the AI response itself is the product page, the comparison guide, and the salesperson. Being cited accurately and favourably in the AI's text is more commercially valuable than ranking #1 in search results, because the user never arrives at your site to see the ranking.

ChatGPT's internal retrieval rank is the single most powerful predictor of citation — pages at position 1 are cited 58% of the time versus 14% at position 10, a 4× gap. Critically, 58% of pages are never cited and 25% are always cited, with only 17% falling between these poles. This bimodal distribution confirms that citation is not a spectrum — it's a threshold. Content that doesn't rank well in ChatGPT's internal retrieval won't be cited regardless of quality. The practical implication: Bing SEO (which predicts ChatGPT retrieval rank) is more important for AI citation than on-page content optimisation alone.

Prompt framing drives an 18x difference in brand mention frequency: a prompt naming a brand generates 14.5 average brand mentions; a category-level query generates 1.68; a generic informational query generates fewer than 1. For brands, this means that being present in the "soft-brand" query space — category-level content that doesn't name a brand but implies one — is the battleground for AI visibility. Brands not appearing in topic-category content lose the majority of AI brand mentions before the user ever types their name.