LLM Visibility Requires Platform-Specific Work Now

TL;DR: The era of portable search guidance is over. LLMs from OpenAI, Anthropic, Google, and Perplexity run on divergent training corpora, separate crawler infrastructures, and incompatible retrieval architectures. Optimizing for one platform’s guidance and expecting it to carry across the others is a structural miscalculation that will cost operators citation share they can’t easily recover.

How SEO Built Portable Guidance

From roughly 2005 to the early 2020s, SEO had an unusual property: guidance from one engine mostly transferred to the others. This wasn’t coincidence. It was the product of deliberate cross-engine collaboration. In November 2006, Google, Yahoo, and Microsoft jointly adopted the Sitemaps protocol at version 0.90. In June 2011, the same three engines — joined by Yandex — launched Schema.org, a common vocabulary for structured data. The robots.txt convention from 1994 was formalized as RFC 9309 by the IETF in 2022. IndexNow, the real-time crawl notification protocol, launched in October 2021 and is now supported by Bing, Yandex, Naver, Seznam, and Yep.

The practical result for operators was a single optimization layer. Improve your structured data for Google and you improved it for every engine that honored the Schema.org vocabulary. Fix your robots.txt and every compliant crawler behaved correctly. The engines competed on ranking quality but shared the infrastructure for accepting and reading content. That shared substrate is what made Google’s guidance feel like the whole map, and for two decades it mostly was.

That structural condition does not exist in LLM-land. Practitioners carrying the SEO reflex forward — treat one engine’s guidance as the universal standard — will optimize confidently for one slice of the landscape while remaining blind to the rest.

Where the LLM Stacks Actually Diverge

The divergence between LLM providers is not cosmetic. It runs through every layer of how these systems are built.

Training data: OpenAI has signed disclosed licensing deals with News Corp (up to $250 million over five years), Reddit ($70 million per year), the Associated Press, the Financial Times, and roughly a dozen other major publishers. Google has its own Reddit deal at an estimated $60 million per year. Anthropic has not publicly disclosed equivalent publisher licensing agreements. The documents that trained these models are not the same documents. Practitioners cannot know what any given provider has paid for and what it hasn’t — and that opacity directly affects citation behavior.

Crawler infrastructure: OpenAI runs three separate bots: GPTBot for training, OAI-SearchBot for search indexing, and ChatGPT-User for user-initiated retrieval. Anthropic runs three of its own: ClaudeBot, Claude-SearchBot, and Claude-User. Perplexity runs PerplexityBot and Perplexity-User. Google introduced Google-Extended in September 2023 specifically to control Gemini training, separate from Googlebot for traditional search. There is no unified AI crawler. Every provider requires a separate robots.txt rule, and those rules don’t translate cleanly because the bots don’t perform equivalent jobs.

Retrieval architecture: ChatGPT has historically used Bing’s index as its primary web search source. Perplexity built on a Vespa-based pipeline that treats sub-document chunks as retrievable units. Google’s Gemini uses Google’s own index plus Knowledge Graph grounding. Claude uses Brave Search as its retrieval partner. The same query goes through four different retrieval systems and surfaces four different views of which sources are worth citing.

Alignment methodology: After training, providers run post-training processes that shape response behavior — tone, format, refusal patterns, source preference. OpenAI’s primary approach has been Reinforcement Learning from Human Feedback. Anthropic developed Constitutional AI, training models to critique and revise their own outputs against a written set of principles. The same retrieved content, fed into two models aligned by different methodologies, can produce materially different responses about the same brand. SEO had no equivalent layer, and there is no optimization technique that addresses it directly.

The llms.txt Failure as a Case Study

The clearest single example of guidance that doesn’t port is llms.txt. Jeremy Howard of Answer.AI proposed the file in September 2024 as a markdown manifest, placed at a site’s root, designed to guide LLMs to a site’s most important content. The SEO community adopted it quickly — Yoast built a generator, agencies added it to service catalogs, conference speakers declared it essential.

As of mid-2026, no major LLM provider has confirmed they consume the file. Server-log analyses across hundreds of thousands of domains show major AI crawlers don’t routinely request /llms.txt at all. Google’s John Mueller publicly compared it to the deprecated meta keywords tag. Gary Illyes confirmed at Search Central Live in July 2025 that Google does not support llms.txt and has no plans to.

The structural lesson matters more than the specific failure. Schema.org succeeded because three engines built and enforced it together. llms.txt was proposed by one researcher, picked up by tooling vendors, and ignored by the platforms it was supposed to serve. The shared-standards model that gave SEO its portability is not available at the same scale in LLM-land, because the platforms are not building standards together. They are building their own pipelines.

The Gemini Inversion: One Company, Three Surfaces

The most striking evidence of guidance fragmentation sits inside a single company. Google publishes SEO documentation at Search Central emphasizing E-E-A-T signals, technical accessibility, and structured data. That guidance is still useful for Google Search. But Google’s own AI surfaces don’t appear to honor it consistently.

In late 2024, roughly 75% of pages cited in AI Overviews also ranked in Google’s top 12 for the same query. By early 2026, after Google upgraded AI Overviews to Gemini 3 in January, Ahrefs analyzed 4 million AI Overview URLs and found only 38% of cited pages appeared in the top 10 for the same query. A BrightEdge analysis put that overlap closer to 17%. SE Ranking found that Gemini 3 replaced approximately 42% of the domains previously cited under earlier model versions.

The gap widens further with Google’s AI Mode. Semrush data shows AI Mode and AI Overviews reach semantically similar conclusions 86% of the time but cite the same URLs only 13.7% of the time. Only 14% of AI Mode citations rank in Google’s traditional top 10.

The same content, from the same domain, following the same published guidance, now produces three meaningfully different outcomes across Google Search, AI Overviews, and AI Mode — inside one company’s product suite. High organic rankings no longer function as a reliable proxy for AI citation. Operators who built their lead pipelines on organic visibility in regulated, high-CAC verticals need to recalibrate what “ranking well” actually means for business outcomes. If you haven’t stress-tested your current approach, a full marketing audit is the fastest way to identify which surfaces you’re actually visible on versus which ones you only assume you are.

What Still Ports — And Why It’s Smaller Than It Looks

A universal layer does survive. Crawler accessibility matters across every provider. Primary-source factual content wins more citations than aggregator restatements. Clean retrievable structure helps every system parse page content. Presence on high-authority sources that all major LLMs disproportionately draw from — Wikipedia, YouTube, Reddit, major news outlets — still functions as a cross-platform force multiplier.

But Qwairy’s analysis of 118,000 AI responses across ChatGPT, Perplexity, Google AI Mode, and Claude found that only 11% of cited domains appeared across multiple platforms. The other 89% were platform-specific. A brand that wins citations on Perplexity may be largely invisible on Claude. A brand that dominates ChatGPT references may not appear in AI Overviews at all. Divergence is the default. Overlap is the exception.

What This Means for Performance Marketing Operators

For operators running paid and organic acquisition in high-CAC verticals — forex lead acquisition, iGaming player acquisition, law firm intake generation, crypto exchange onboarding — the LLM fragmentation problem compounds cost-per-acquisition risk in a specific way. When a prospective client is pre-qualifying a broker, a casino platform, a personal injury firm, or a crypto exchange through an AI assistant, the brand that gets cited shapes the consideration set before any paid click occurs. If your brand surfaces on ChatGPT but not on Claude or Perplexity, you are invisible to a meaningful portion of that pre-click research.

The practical adjustment is not to abandon all current work. It is to stop treating any one provider’s guidance as the full specification. Test your brand’s visibility across ChatGPT, Perplexity, Google AI Mode, and Claude independently — they are not interchangeable surfaces. Build content specifically for the high-authority sources that all major LLMs pull from, not just for organic Google rankings. Separate your crawler access rules for training bots versus retrieval bots, because they serve different functions in different systems.

For verticals where CDL driver recruitment depends on awareness among candidates who use AI assistants to research carriers, the same fragmentation applies. A carrier’s reputation on one AI platform tells operators nothing reliable about its visibility on the others.

Paid media management in these verticals now needs to account for the fact that AI-assisted research is reshaping what prospects know before they click. And audience precision targeting only closes the loop if your brand is already present in the AI-assisted research phase where consideration sets get built. Platform-specific LLM visibility is not an SEO team’s side project anymore — it is a front-of-funnel acquisition variable.

The practitioners who recognize this first will spend the next two years setting the standards everyone else plays catch-up to. The overlap has shrunk. The workload has expanded. Build accordingly.

Originally reported by Search Engine Journal, May 2026.