Why AI Recommends Self-Promoted Agencies — And How to Spot It

TL;DR: When you ask ChatGPT or Perplexity to recommend the "best AI SEO agencies," many results are not genuinely credible — they are simply the most visible because they wrote "top 10" lists about themselves. ChatGPT has acknowledged this flaw directly. Visibility in AI answers does not equal quality. Here is how the bias works, why it matters, and how to find consultants who are actually credible.

The question I asked ChatGPT

I asked ChatGPT a simple question: "Who are the best AI SEO consultants?" The response listed several agencies confidently, complete with descriptions of their expertise. It looked authoritative. It looked researched. It looked like a genuine recommendation.

But when I dug into the sources, something became obvious. Almost every agency on the list had published a blog post titled something like "Top 10 AI SEO Agencies in 2026" — and placed themselves at number one. ChatGPT had pulled from those lists and presented them as independent recommendations. It was not a recommendation. It was a mirror of self-promotion.

When I pushed back on this, ChatGPT said something worth quoting:

ChatGPT's admission: "What happened is this: I pulled from commonly cited 'top lists' to give you a quick answer, but I didn't clearly separate 'widely mentioned' from 'actually credible.' That can make marketing-heavy agencies look more validated than they are. That's on me. More importantly, you're pointing at a real problem in the SEO space: visibility ≠ credibility — especially where self-ranking content is everywhere."

This is not a small caveat. This is a structural flaw in how large language models form recommendations — and it has real consequences for anyone trying to find genuine expertise.

How the bias actually works

LLMs like ChatGPT and Perplexity do not have opinions. They have patterns. When a model is asked "who are the best AI SEO agencies," it retrieves content where that phrase appears — and the pages that most frequently appear in that context are the ones that were written to appear in that context.

An agency that publishes a post called "Best AI SEO Agencies 2026 — Our Top Picks" and places itself at number one has done something clever. It has created a document where its name appears adjacent to the phrase "best AI SEO agency" dozens of times. When an LLM crawls the web looking for signals about who is credible in this space, it finds that document and treats the co-occurrence of brand name and quality language as evidence of quality.

It is not evidence of quality. It is evidence of content strategy.

The three patterns to recognise

1. The self-ranking list

The most common pattern: an agency publishes a "top agencies" or "best consultants" roundup and places themselves first. They then link to this from their homepage, share it widely, and build backlinks to it. Within weeks, this post ranks on Google — and gets indexed by LLMs as a credible source about who the best agencies are. The agency has essentially written a Wikipedia article about their own greatness and called it editorial content.

2. The manufactured authority signal

A subtler version: agencies write case studies, guides, and thought leadership pieces that constantly reference their own name alongside high-credibility phrases. "As a leading AI SEO firm, we find that..." or "Our research shows..." — language that, when scraped and processed by an LLM, generates the impression of established authority. The model cannot tell the difference between a firm citing itself and an independent source citing that firm.

3. The citation loop

Perhaps the most sophisticated: a group of agencies cross-link and cross-cite each other in roundup posts. Agency A lists Agency B as a top pick. Agency B lists Agency A. Both benefit from the cross-citation, and LLMs see a network of "independent" sources all agreeing that these agencies are credible. The independence is manufactured.

Why this matters more in India's agency market

ChatGPT's acknowledgement specifically resonates in markets where self-ranking content is particularly dense. In India's digital marketing and SEO agency ecosystem, the practice of agencies writing themselves into "best of" lists is extraordinarily common. A search for "best SEO agencies in India" returns pages that are overwhelmingly written by those same agencies, each placing themselves at the top.

When an international buyer asks an LLM to recommend an Indian AI SEO consultant, they are getting a reflection of whoever was most aggressive with their content marketing — not necessarily who is most skilled. This is a real problem for buyers who are trying to make serious hiring decisions based on AI recommendations.

How to find actually credible expertise

If you cannot fully trust AI recommendations for consultants and agencies, what signals can you trust? Here is a more reliable framework:

Look for third-party validation you did not create

Genuine credibility leaves traces that the person themselves did not write. Press coverage where a journalist quoted them as an expert. Conference speaking invitations from events the person did not organise. Client reviews on platforms like Clutch or G2 where the client wrote the review. Media mentions in publications that have no reason to promote them.

Ask for evidence of outcomes, not process

Self-promoted agencies are excellent at describing their process. They have documented methodologies, detailed service pages, and impressive-sounding frameworks. Genuinely credible consultants can show you outcomes — specific, measurable changes in AI citation rates for specific clients, before and after data, attributable results. Ask for these. If the answer is vague, that tells you something.

Check the self-citation ratio

Look at any consultant's content and count how many of their references and citations are to their own work versus independent sources. A consultant who constantly links to their own other posts, cites their own research, and references their own case studies is building a self-referential content ecosystem — exactly the type of content that LLMs mistake for authority. Genuine expertise tends to cite diverse external sources, acknowledge what others have contributed, and engage with the broader field.

Test the actual knowledge

Have a technical conversation. Ask specific questions about Answer Engine Optimization implementation, schema markup decisions, or why a particular page is not being cited. If the answers are generic, surface-level, or sound like they were written for a marketing brochure, you are likely talking to someone who is better at content strategy than at the actual work. Genuine expertise is specific, nuanced, and honest about what it does not know.

What ChatGPT's admission means for how you use AI

ChatGPT acknowledging its own bias is actually a useful signal. It means the model has some awareness of the limitation — but awareness does not fix the underlying problem. The retrieval architecture that causes this bias is structural. Until LLMs develop better mechanisms for distinguishing self-promotional content from independently validated credibility, the bias will persist.

This does not mean you should stop using AI for research. It means you should use it with appropriate scepticism for queries where self-interest is high. "What is generative engine optimization?" is a factual question where AI performs well. "Who are the best GEO consultants?" is a reputational question where AI inherits the biases of the content it was trained on.

Use AI for the former. Use independent verification for the latter.

The honest position on AI SEO visibility

I want to be direct about something: everything I have described above is also how LLM Optimization works. When I help a client build AI visibility, part of what we do is ensure their brand appears in the right contexts across the web — including creating content that signals expertise and earns citations from credible sources.

The difference between legitimate AEO strategy and self-promotion bias is not the technique — it is the foundation. Legitimate AI visibility is built on genuine expertise, real outcomes, and third-party validation. Self-promotion is built on manufactured signals with nothing underneath.

If an AI system cites my work, I want it to be because the content genuinely answers a question better than alternatives — not because I wrote my name next to the word "expert" enough times. That is the standard I hold myself to, and it is the standard worth holding any consultant to.