TL;DR
- ChatGPT and Perplexity use fundamentally different citation mechanisms — Perplexity runs live web searches, ChatGPT draws from training data, which means the same query returns completely different results
- The agencies currently dominating GEO citation lists got there by writing their own roundup articles and appearing on them — a self-reinforcing loop
- Citation rates vary dramatically by platform: Perplexity cites businesses in nearly every response, Google AI Overviews in about a third, ChatGPT in roughly one in six
- For buyer queries with local or specific intent, no credible small or mid-sized business has yet established a dominant citation position — the window is still open
Clever Merchants ran a citation analysis in May 2026 to answer two questions: who is currently being cited by AI search tools for queries in our category, and whether we are among them.
The short answer to the second question is no, not yet. This is expected for a site launched in early 2026. What was more interesting was what the research revealed about how citations are being won — and how mechanical the current process actually is.
The test: 40 queries across four AI tools
We ran queries through ChatGPT (with and without web search), Perplexity (with live web search), and Google Gemini, covering four query categories:
- GEO agency queries — “best generative engine optimization agency”, “top GEO agencies 2026”, “which agency gets businesses cited in ChatGPT”
- AI SEO queries — “best AI SEO agency”, “who are the best GEO agencies for AI search visibility”
- Methodology queries — “how do I get my business cited in ChatGPT”, “what is generative engine optimization”
- Competitor and brand queries — direct name lookups and category queries for specific agencies
Here is what the data showed.
Finding 1: Perplexity and ChatGPT are entirely different citation systems
This is the most important structural insight from the research, and the one most businesses and even most agencies miss.
ChatGPT (without web search) generates answers from training data. If your business is not represented in ChatGPT’s training corpus — through web pages, directories, reviews, and publications that were indexed before the training cutoff — you will not appear. The training cutoff also means that very recent businesses and websites are effectively invisible to base ChatGPT.
ChatGPT with web search (GPT-4o in browse mode) is a different system. It fetches live web results and synthesises them. In this mode, schema markup and freshly indexed content matter significantly. A well-structured site can appear in web search-enhanced ChatGPT responses faster than in base ChatGPT.
Perplexity runs live web searches for almost every query. It cites sources explicitly and indexes new content fast. According to DataForSEO’s citation analysis data, Perplexity cites businesses in approximately 97% of its responses, compared to roughly 34% for Google AI Overviews and 16% for base ChatGPT.
Practical implication: Perplexity is where new businesses can earn citations fastest. If your schema, content, and entity data are clean and indexed, Perplexity can begin citing you within days. ChatGPT training data citations take much longer but are reinforced by web search for current users.
Finding 2: The agencies dominating GEO citation lists got there by writing their own lists
When Perplexity was asked “what are the best GEO agencies in 2026?”, it returned a ranked list with scores: Linkflow (9.4/10), Rock The Rankings, First Page Sage (8.4/10), Siege Media (8.7/10), iPullRank (8.6/10), Omniscient Digital (8.2/10).
The sources Perplexity cited for this list: a blog post on linkflow.ai titled “Top GEO SEO Agencies”, a blog post on rocktherankings.com titled “Top 13 GEO Agencies for Q1 2026”, a blog post on thedigitalelevator.com, and a post on mintcopywritingstudios.com.
The pattern: Linkflow appears in Perplexity’s top position because Linkflow published a “top GEO agencies” article that included themselves at the top. Rock The Rankings appears because Rock The Rankings published a “top 13 GEO agencies” article that included themselves. Perplexity found those articles in web search, treated them as authoritative sources, and synthesised the rankings.
This is not unique to GEO. It is how Perplexity handles most “best of” queries in emerging categories where there is no independent authority: it reads what ranks in Google, and what ranks in Google is often written by the businesses that stand to benefit from being in the list.
The implication for Clever Merchants is specific and actionable: getting listed in third-party roundup articles on sites that Google ranks for GEO agency queries is the fastest path to Perplexity citations. Being on those lists matters more right now than the content of our own site.
Finding 3: ChatGPT does not yet recognise “generative engine optimization” as a distinct category
When asked “best generative engine optimization agency” without web search, ChatGPT responded with confusion about what the query meant, then defaulted to listing traditional AI-enhanced SEO agencies: MarketMuse, Semrush Agency Partners, Frase, Victorious SEO, Neil Patel Digital. None of these businesses offer GEO in the sense of optimising for AI citations. They use AI tools for traditional SEO.
When asked “best GEO agencies for AI search visibility”, ChatGPT interpreted “GEO” as geographic targeting and listed local SEO agencies: BrightEdge, Semrush, LocaliQ, Ignite Visibility, Victorious SEO.
This matters for two reasons. First, it means there is currently no agency strongly associated with “generative engine optimization” in ChatGPT’s training data — not Linkflow, not First Page Sage, not Clever Merchants. The category is new enough that the training corpus has not yet consolidated around clear leaders.
Second, it means the content strategy for GEO agencies right now should be definitional. Articles that clearly define what GEO is, what distinguishes it from traditional SEO, and what GEO implementation actually involves are the content types most likely to be indexed and cited as the category matures in AI training data. The blog post you are reading is doing exactly that.
Finding 4: For specific buyer intent queries, no one has established dominance
When we tested high-intent buyer queries — “which agency can get my business cited in ChatGPT”, “how do I get my service business to appear in AI search answers” — ChatGPT with web search returned explanatory content rather than agency recommendations. It described the method (schema, directories, GBP, content) but did not name specific agencies.
This is the most commercially significant gap in the research. The queries that a buyer types immediately before making a purchase decision — not “what is GEO” but “which agency does GEO and can I trust them” — are currently returning generic how-to answers rather than citing specific providers.
For any agency with a well-structured service page, detailed methodology content, and an established citation footprint, these are the queries to target. The position is open.
What this means for businesses (not just agencies)
The research above focuses on the GEO agency category because that is Clever Merchants’ own commercial context. But the underlying mechanics apply to every category in every sector.
The businesses currently being cited in AI tools for specific product or service queries got there through a combination of training data density (they have been mentioned consistently across many sources for a long time) and structured signals (schema, GBP, question-answering content). In most local and niche categories, those positions are not yet locked in.
The window is clearest in categories where:
- The buyer query is specific (“best solar installer Cork”, “commercial roofing contractor Dublin”)
- The category is genuinely local or specialist (fewer national brand competitors)
- The existing competitors have not implemented GEO signals
For businesses in those categories, the implementation investment required to move from invisible to cited is relatively small. The Shamrock Electrical result — cited within 72 hours — happened because no other Irish electrical wholesaler had done the structural work. The query had a gap. We filled it.
The methodology behind this research
All queries were run through:
- ChatGPT (GPT-4o, no web search) via the DataForSEO ChatGPT scraper
- ChatGPT (GPT-4o with web search) via the DataForSEO LLM response tool
- Perplexity (Sonar model with web search enabled) via the DataForSEO LLM response tool
Queries were not cherry-picked for results. The full set covered agency recommendation queries, methodology queries, and brand-specific queries. Results were taken from the first response with no prompt engineering.
Citation counts and platform comparison statistics referenced in this article are drawn from DataForSEO’s AI citation analysis data and third-party research published by First Page Sage and Siftly.
Frequently asked questions
Why does ChatGPT give different answers to the same query each time? ChatGPT generates probabilistic responses — the same query can produce different outputs because the model samples from a distribution rather than returning a fixed answer. This is why citation research requires running queries multiple times and looking for consistency, not single screenshots.
Does being cited by Perplexity help with ChatGPT citations? Indirectly. Perplexity citations mean your content is being indexed and retrieved in live web searches. That content, as it accumulates mentions and links, strengthens the corpus that future ChatGPT training runs will draw on. The fastest path to durable ChatGPT training citations is building a strong Perplexity citation record first.
How do we know which queries to test for our own business? Start with the questions your buyers actually ask before they contact you. “Best [service] in [location]”, “how to choose a [service provider]”, “what does [service] cost” — these are the queries that drive citations with commercial intent. The free AI Visibility Check runs exactly these queries against your business category and location and tells you where you stand.
Is the GEO agency category too crowded to enter now? No. The category is new and the agencies currently being cited are large US firms that have pivoted from traditional SEO. They are not specialists. The niche for a specialist GEO agency focused on specific business types — trades, ecommerce, professional services — is not currently occupied by anyone who has both the credentials and the citation footprint to hold it.
What is the single most impactful thing a business can do today? Add FAQPage schema to your main service or product pages and publish four to six questions your buyers ask, with direct answers. This is the highest-leverage single action in GEO. It makes your content machine-readable for AI systems that pull verbatim answers and simultaneously makes you eligible for Google’s FAQ rich results. It can be implemented on most CMS platforms in under two hours.
