There is no ranking in ChatGPT, Perplexity, or Gemini. Not one of the three publishes a ranking algorithm, two of them do not query the same index, and what each vendor does publish is documentation about retrieval.
That gap between what firms are told and what the vendors have written down is why a practice can sit at position one on Google and never surface in a ChatGPT answer. We work from those vendor documents at Lexicon Legal Content rather than from other agencies’ summaries of them, and the three engines diverge further than the marketing around them admits.
ChatGPT, Perplexity and Google run on three separate indexes served by six separate crawlers, so there is no single ranking for a law firm to optimize toward.
Each vendor publishes retrieval documentation rather than a ranking algorithm, and those documents contradict a good deal of the advice sold to law firms.
ChatGPT Runs on Someone Else’s Index
OpenAI does not operate a search index of its own for ChatGPT search. Its documentation states that ChatGPT search leverages third-party search providers along with content supplied directly by partners, and OpenAI’s help documentation names Bing as one of those providers.
OpenAI publishes four separate bots, and the difference between two of them corrects the most common mistake in law firm marketing. GPTBot is described as crawling content that may be used in training foundation models, while OAI-SearchBot is described as surfacing websites in ChatGPT’s search features.
Blocking GPTBot therefore does not remove your firm from ChatGPT search. Plenty of firms have been told otherwise, and a few have blocked the wrong agent on that advice, which is a costly way to lose the visibility inside ChatGPT results they were paying to build.
Perplexity Scores Passages, Not Pages
Perplexity runs its own index and evaluates your content in fragments. Its engineering documentation describes an index covering hundreds of billions of webpages, and states that its retrieval infrastructure divides documents into fine-grained units that are individually surfaced and scored against the query.
That single sentence is the only vendor-documented argument for answer-first writing that exists across any of the three engines. If the scoring unit is the passage, a self-contained paragraph naming the jurisdiction and the standard is the asset, and the page around it is packaging.
Perplexity’s crawler behavior also differs from OpenAI’s in a way worth noting before anyone edits robots.txt. PerplexityBot is documented as not being used to crawl content for foundation models at all, so there is no training-versus-search split to manage.
Google Wants Nothing Special From You
Google’s position is that AI Overviews and AI Mode require no separate optimization. Its documentation states there are no additional requirements to appear in either surface and no other special optimizations necessary, and that a page must be indexed and eligible to appear with a snippet.
The mythbusting in Google’s guidance on helpful content and its generative AI guide kills several paid recommendations at once. No AI text files, because Google Search does not use them. No required structured data. No requirement to fragment your content.
One line settles the argument firms have been having since Google-Extended appeared. Google states that Google-Extended does not affect a site’s inclusion in Google Search and is not used as a ranking signal, which means blocking it protects nothing about your AI Overview presence and costs nothing either.
What Actually Differs, and What It Changes
Three architectures produce three different jobs. ChatGPT depends on an index it does not own, Perplexity owns its index and scores passages inside it, and Google runs retrieval over the same ranking systems that already decide your organic position.
The practical consequences separate cleanly. Bing indexation matters for ChatGPT, though OpenAI has never stated it as a requirement and anyone claiming otherwise is inferring. Passage-level structure matters most for Perplexity. For Google, ordinary indexing and snippet eligibility are the whole technical story.
There is a shared failure mode underneath all three. Vercel and Merj measured AI crawler behavior in December 2024 and reported that none of the major AI crawlers rendered JavaScript, which means the client-rendered attorney bios, accordion practice area pages, and review widgets common on law firm sites are invisible to two of these three engines.
That makes a firm site’s structure a retrieval question rather than a design one. Orphaned pages and content that only exists after a script runs are content these crawlers never see.
The Legal Wrinkle Nobody Gets Right
Legal content is treated as high-stakes, but not the way the industry says. Google’s Search Quality Rater Guidelines name four categories in the your money or your life section, being health or safety, financial security, government and civics, and other, and “legal” is not one of them.
Legal content lands inside those definitions rather than beside them. An injury or bankruptcy matter is financial security; an immigration or criminal matter is government and civics. Saying that precisely is more accurate than the shorthand every competing page uses.
Authorship is the one signal with real documentation behind it, and only from Google, whose guidance asks whether it is self-evident who authored the content and whether bylines lead to further information about the author. OpenAI and Perplexity publish nothing equivalent, so attorney bylines are a documented Google signal and an educated guess everywhere else.
Building verifiable credibility signals is therefore worth doing for Google’s stated reasons rather than for a claim about ChatGPT that no vendor has made. The work is the same either way.
Frequently Asked Questions
Does ChatGPT use Google’s rankings?
No. OpenAI documents that ChatGPT search draws on third-party search providers and content supplied by partners, naming Bing among those providers. Your position in Google’s organic results carries no documented weight inside ChatGPT’s own retrieval process.
How long does it take to show up?
The answer differs by engine. Perplexity states its index processes tens of thousands of update requests per second, Google follows its ordinary crawl and index cycle, and OpenAI documents no timeline because it depends on another company’s index.
Should we block AI crawlers?
The evidence is genuinely unresolved. Blocking a training crawler has not reliably removed sites from citations in published testing, while blocking a retrieval crawler correlates with near-zero citations in the one study that normalized for site prominence.
Give Your Developer Something Specific
Almost none of this is a marketing task. It is a list for whoever maintains the site: confirm OAI-SearchBot and PerplexityBot can reach the pages, verify Googlebot indexing and snippet eligibility, and identify the content that only exists once a script has run.
We produce that list, line by line, in a form an engineer can work through. Anything we then write to fill the gaps is checked by a licensed attorney before it ships, so the fix does not create a bar problem while solving a crawler one.
Once access is sorted, the metrics worth tracking split by engine, and we set that up in the same pass. Call 877-486-8123 and ask for the crawler check, or take a free sample first.
David Arato, JD is the founder of Lexicon Legal Content, an attorney-owned legal content marketing agency serving law firms since 2012. He developed Lexicon’s approach to SEO, AEO, and GEO around the credibility, clarity, and consistency signals that can strengthen a firm’s visibility across search and AI-driven platforms. David is a frequent contributor to Attorney at Law Magazine and Attorney at Work and a recurring guest on legal marketing podcasts.
