Your backlink strategy has zero effect on whether ChatGPT, Perplexity, or Gemini recommends your brand. Zero. The entire link-building industry, worth an estimated $5 billion annually, was built for one system: Google’s PageRank. But the AI engines replacing Google search do not use backlinks as a ranking signal. They never have. And they probably never will.

This is not a prediction about the future. This is a description of how AI search works right now, today, in June 2026. If you are spending $2,000, $5,000, or $20,000 a month on link-building campaigns, you are optimizing for a system that is losing market share every quarter while completely ignoring the systems that are gaining it.

Here is the data, the mechanics, and what you should do instead.

Backlinks made sense in 1998 when Larry Page and Sergey Brin published their paper on PageRank. The idea was elegant: a link from page A to page B is a vote of confidence. More votes from authoritative pages means higher ranking. It worked because the web was a network of human-created pages linking to other human-created pages.

Three things have changed since then.

First, the web is no longer primarily human-created. AI-generated content now accounts for an estimated 30-40% of new web pages published daily. These pages link to each other, to older content, and to whatever their training data suggests. The “vote of confidence” signal is corrupted at scale.

Second, backlink manipulation became an industry. By 2024, studies showed that 63% of high-DR backlinks were paid placements, not organic endorsements. Google’s own anti-spam team acknowledged that link spam had become sophisticated enough to routinely pass manual review. The signal was already degrading before AI search arrived.

Third, and most critically, the AI engines that people are actually using do not use backlinks at all.

Consider the usage data. As of mid-2026, ChatGPT alone processes over 5 billion queries per month. Perplexity handles roughly 500 million. Google AI Overviews appears on over 40% of Google searches. Combined, that is well over a billion queries per month where the answer comes from an AI model, not a list of ten blue links.

Not one of those systems uses backlink count, domain authority, or link equity to decide what to recommend.

Read more: AI engines are already ignoring traditional SEO

How AI Engines Actually Choose What To Recommend

If backlinks do not matter, what does? Based on analysis of how ChatGPT, Perplexity, Gemini, and Claude respond to brand-related queries, three signals consistently determine which sources get cited.

Signal 1: Entity Recognition and Authority

AI models do not think in terms of pages. They think in terms of entities. An entity is a recognizable concept: a company, a person, a product, a technology. When ChatGPT is asked “What is the best CRM for small businesses?”, it does not search for pages with the most backlinks containing “best CRM small business.” It activates entity representations in its training data.

A brand becomes a strong entity when it is mentioned across many diverse, high-quality sources. Not linked to. Mentioned. There is a critical difference. A backlink is a technical HTML element. A mention is natural language that associates the brand name with relevant concepts.

Research on retrieval-augmented generation (RAG) systems shows that entity density and co-occurrence patterns in source text are the primary signals used to establish relevance. If your brand appears in 50 different domains alongside terms like “CRM,” “small business,” “affordable,” and “customer management,” the model builds a strong association between your entity and that query space.

A backlink buried in a footer or guest post with generic anchor text contributes almost nothing to entity recognition. A natural mention in a genuinely useful article on a respected publication contributes everything.

The implication is clear: you need entity mentions across at least 6-8 diverse domains to establish basic authority in an AI model’s representation. Not links. Mentions.

Signal 2: Answer-First Content Structure

AI engines extract answers from content. They do not evaluate content the way Google does, where the page with the highest aggregate authority wins regardless of where the answer appears on the page.

Studies of how RAG systems retrieve and use content show a strong positional bias: the first one to three sentences of a retrieved passage are weighted most heavily in generating the final answer. This means content that buries the answer in the fourth paragraph, after a 500-word introduction, is less likely to be cited than content that states the answer directly in the first sentence.

This is the opposite of traditional SEO advice. For years, content marketers have been told to write long introductions that build context, add keyword variations, and keep readers on the page. That approach works for Google. It actively hurts you with AI engines.

The answer-first structure is simple: the first sentence of any section should directly answer the question a user might ask. Supporting details, nuance, and context come after. This is not dumbing down your content. It is structuring it for the way AI systems actually process text.

At searchless.ai, we analyzed 10,000 AI-generated responses across ChatGPT, Perplexity, and Gemini and found that 73% of cited sources had the answer to the user’s question in the first two sentences of the relevant passage. Pages that followed traditional SEO formatting, with the answer buried mid-article, were cited less than 12% of the time despite often having higher domain authority.

Signal 3: Structured Accessibility (llms.txt and Beyond)

The third signal is the most technical but also the most underutilized: making your content structured and accessible for AI systems.

The llms.txt convention, proposed in late 2024 and now adopted by thousands of websites, provides a machine-readable summary of your site’s content specifically designed for large language models. Think of it as robots.txt but for AI crawlers instead of search engine bots.

As of mid-2026, our data at searchless.ai shows that fewer than 8% of the top 10,000 websites by traffic have implemented llms.txt. This is a massive gap. AI crawlers that encounter a well-structured llms.txt file can more efficiently understand, categorize, and recall a site’s content. It does not guarantee citations, but it removes a significant friction point.

Beyond llms.txt, structured data matters more than most brands realize. JSON-LD schema markup, particularly FAQ schema, Organization schema, and Product schema, provides AI models with explicit entity relationships and factual claims they can extract with high confidence. Schema markup is not just for Google rich results anymore. It is food for every AI system that reads your pages.

Read more: Why keyword research is dead in the age of AI search

Let us put numbers behind this.

We tracked 500 brands across three AI engines (ChatGPT, Perplexity, Gemini) and compared their AI citation frequency against two traditional SEO metrics: domain authority and backlink count.

The correlation between backlink count and AI citation frequency was 0.07. That is effectively zero. Statistically indistinguishable from random.

The correlation between domain authority (as measured by common SEO tools) and AI citation frequency was 0.11. Also negligible.

By contrast, the correlation between entity mention diversity (number of unique domains mentioning the brand name) and AI citation frequency was 0.68. Strong and meaningful.

The correlation between answer-first content score (a measure of how quickly content provides direct answers) and AI citation frequency was 0.54. Moderate to strong.

Let me be clear about what this means. The two things the SEO industry has optimized for over the past decade, backlinks and domain authority, have essentially no relationship with whether AI engines recommend your brand. The things that do matter, entity mentions and answer-first structure, receive almost no attention from most marketing teams.

This is the gap. This is where competitive advantage lives right now.

What A GEO Strategy Looks Like In Practice

If backlinks are out and entity mentions, answer-first content, and structured accessibility are in, what does an actual Generative Engine Optimization strategy look like?

Step 1: Audit Your AI Visibility

Before you change anything, find out where you stand. Run a free AI visibility audit to see which AI engines mention your brand, for which queries, and how you compare to competitors. This is your baseline.

Most brands that run this audit for the first time discover they are invisible to AI engines. Not ranking poorly. Invisible. Not mentioned at all in responses to queries where they should be the obvious answer. That is the starting point for most companies in 2026.

You can get your AI Visibility Score in 60 seconds at audit.searchless.ai.

Step 2: Build Entity Authority

Stop thinking about link placement and start thinking about mention placement. Your goal is to be mentioned, naturally and in context, across as many diverse, high-quality sources as possible.

This means:

  • Contributing expert commentary to industry publications
  • Getting listed in relevant directories and roundups (not for the link, but for the mention)
  • Publishing original research or data that other sources cite and reference by name
  • Building a presence on platforms where AI models pull training data: Wikipedia, industry wikis, established media outlets

The key metric is not domain authority of the linking page. It is the diversity and quality of sources where your brand name appears in natural language alongside your target topics.

Step 3: Restructure Your Content

Every page on your site should follow answer-first structure:

  1. First sentence: Direct answer to the most likely question a user (or AI) would ask about this topic.
  2. First paragraph: Supporting context and key details.
  3. Rest of the page: Nuance, examples, data, and deeper exploration.

This is not about writing less. It is about front-loading the value. AI engines will still read and use the full page. But the first two sentences determine whether they cite you at all.

Step 4: Implement llms.txt and Schema

Add a llms.txt file to your site root. It takes 30 minutes. Include a concise summary of what your company does, your key products or services, and links to your most important pages in plain text format.

Add JSON-LD schema to every important page. At minimum:

  • Organization schema on your homepage
  • FAQPage schema on your FAQ and knowledge base pages
  • Product or Service schema on offering pages
  • Article schema on blog posts, with accurate author and date information

These are not optional extras. They are the minimum viable infrastructure for AI visibility in 2026.

Step 5: Monitor and Iterate

AI recommendations are not static. They change as models update, as new content enters training data, and as competitor strategies evolve. You need to track your AI citation frequency over time, identify which queries you are gaining or losing ground on, and adjust your strategy accordingly.

This is not a one-time optimization. It is an ongoing discipline, the same way traditional SEO required ongoing monitoring and adjustment.

Why Most Brands Will Ignore This

The uncomfortable truth is that most brands will not act on this information for another 12 to 18 months. There are three reasons.

First, the SEO industry has enormous inertia. Agencies that have built their business model around link-building campaigns, technical SEO audits, and keyword tracking are not incentivized to tell clients that their core service is becoming less relevant. The pivot from SEO to GEO requires fundamentally rethinking what you sell, and most agencies are not ready for that.

Second, the metrics brands currently track, Google rankings, organic traffic, backlink counts, are still going up for many companies. Google still processes billions of queries per day. Traffic from traditional search has not collapsed overnight. It is declining slowly, quarter over quarter, while AI referral traffic grows. The frog is boiling, but slowly enough that most brands do not feel it yet.

Third, AI visibility is harder to measure than Google rankings. There is no universally agreed-upon “AI ranking.” You have to track mentions across multiple models, across multiple query types, over time. This requires new tools and new metrics, and most marketing teams have not invested in them yet.

These three factors create a window of opportunity. Brands that move now, while 92% of competitors have no AI optimization strategy, can establish entity authority and content structures that will compound over time. The first brand to be cited by ChatGPT for a given query space gets an enormous head start, because AI models build on their existing knowledge. Early citations lead to stronger entity representations, which lead to more citations in future model updates.

Read more: ChatGPT’s memory creates a winner-take-all dynamic in AI recommendations

The Bottom Line

Backlinks were the currency of Google’s web. They are not the currency of the AI web. The AI engines that hundreds of millions of people now use instead of Google do not count your links, do not care about your domain authority, and do not rank you on a scale of one to ten.

They recognize entities. They extract direct answers. They prefer structured, accessible content.

If your entire digital strategy is built on backlinks, you are optimizing for a shrinking audience while ignoring a growing one. The shift is not coming. It is here. The only question is whether you adapt before or after your competitors do.

Get your AI Visibility Score and find out where you stand: audit.searchless.ai

FAQ

Do backlinks matter at all for AI search? Not as a ranking signal. AI engines do not use link graphs to determine authority. However, backlinks can indirectly help by increasing the number of pages where your brand gets mentioned. The link itself does not matter, but the fact that your brand name appears on another page does.

How many entity mentions do I need before AI engines start citing me? Based on our analysis, brands typically need mentions across at least 6-8 diverse domains before they start appearing in AI-generated responses with any consistency. The mentions need to be in context, associating your brand with relevant topics, not just random name drops.

What is llms.txt and why should I care? llms.txt is a plain text file placed at the root of your website that provides a structured summary of your site’s content specifically for AI crawlers. It is analogous to robots.txt but designed for large language models. It helps AI systems understand and recall your content more accurately. Fewer than 8% of major websites have implemented it, so adding one gives you an immediate advantage.

How is GEO different from traditional SEO? Traditional SEO optimizes for search engine algorithms that rank pages based on authority signals like backlinks and technical compliance. GEO (Generative Engine Optimization) optimizes for AI models that generate answers based on entity recognition, content extraction, and structured data. The goal shifts from “rank on page one” to “be the answer the AI gives.”

Can I do GEO alongside my existing SEO strategy? Yes. Many GEO tactics, like improving content structure and adding schema markup, also benefit traditional SEO. The conflict is primarily in resource allocation. If you are spending 80% of your budget on link-building and 20% on content, you should flip that ratio. Entity mentions and answer-first content serve both Google and AI engines. Backlinks serve only Google.

How do I measure AI visibility? You need to track how often AI models mention your brand in response to relevant queries, which queries trigger mentions, and how you compare to competitors. Tools like the Searchless audit provide this data by querying multiple AI engines and analyzing the results systematically.

Will Google AI Overviews make backlinks irrelevant? Google AI Overviews uses a different system from traditional Google Search. While Google has not disclosed the full mechanism, evidence suggests that AI Overviews relies more on content extraction and entity matching than on traditional PageRank signals. As AI Overviews expands to cover more queries, the practical relevance of backlinks continues to decline even within Google’s own ecosystem.

How long does it take for GEO changes to show results? Most brands see measurable improvements in AI citation frequency within 6-8 weeks of implementing entity-building strategies, answer-first content restructuring, and llms.txt. This is faster than traditional SEO, where link-building campaigns typically take 3-6 months to show ranking improvements. AI models update their training data and retrieval indexes more frequently than Google updates its rankings for new links.