What Is the Citation Gap? Why AI Doesn't Know Your Business

Last updated: 2026-10-12

TL;DR

Definition: The Citation Gap is the space between what AI says about your brand and what your business actually is.

Where buyers start: In G2’s August 2025 survey of 1,000+ B2B software buyers, half said they now start the buying journey in an AI chatbot instead of Google Search [1].

Whose pages AI reads: A brand’s own website supplies only about 5 to 10 percent of the sources AI search references, per McKinsey [2].

Our own numbers: In our August 2026 audit of past TwoSetAI guests, 95 percent appeared on a Google surface and 86 percent were cited by at least one AI engine [3]. When ChatGPT did not cite a YouTube video, it reached guests through third-party authoritative posts (founder profiles, twosetai.com pages, Medium posts), which is why every TwoSetAI episode ships with a written editorial layer alongside the recording [3].

Half of the B2B software buyers G2 surveyed in August 2025 said they now start the buying journey in an AI chatbot instead of Google Search [1]. So here is a question worth two minutes: if a buyer, an investor, or a senior candidate asked ChatGPT, Perplexity, Google AI Mode, or Claude about your company this morning, what did they read?

You probably don’t know. McKinsey found that only 16 percent of brands systematically track their performance in AI search [2]. This series is built around one question: why doesn’t AI know your business well enough to recommend you? I’ll answer it in four parts: what the gap is, why it opens, how to measure yours, and what closes it. I’m writing for pre-Series B AI founders, whose first impression is now written by a system they can’t see.

What is the Citation Gap?

The Citation Gap is the space between what AI says about your brand and what your business actually is. It takes three shapes:

Absent. The engines don’t mention you, so you never enter the consideration set.
Wrong. They mention you with a stale description, the wrong category, or a competitor’s framing.
Mis-tiered. They mention you as a footnote when you are the specialist. I call the tier an engine puts you in the Buyer Bucket. It is set by what the engine can cite about you, not by what your business is.

This is not the same as being indexed. Google says a page must be indexed and snippet-eligible to appear as a supporting link in AI Overviews and AI Mode, with no additional technical requirements [4]. Indexing gets you into the building. Whether an engine names you is a separate outcome, which I cover in The Difference Between Being Indexed and Being Cited by AI.

I also use a term for the moment this matters. The Counterparty Check is the quiet lookup a buyer, investor, or candidate runs on you in an AI assistant before agreeing to talk. You never see it. You only see the call, and whether the other person already has the wrong idea.

Why do buyers, investors, and candidates check AI before they take your meeting?

It has become a first stop, and several independent sources agree.

G2: half of surveyed B2B software buyers start in an AI chatbot, a 71 percent jump from its survey four months earlier [1].
Forrester: in research analyst Christina Schmitt cited in June 2026, 87 percent of B2B buyers selected a “generative AI conversational search tool” as a meaningful interaction, the top-ranked option (the post is tagged EMEA and gives no sample size) [5].
McKinsey: half of 1,927 US consumers surveyed in August 2025 intentionally use AI-powered search, and 44 percent of those users call it their primary source, ahead of traditional search at 31 percent [2].
Pew: in March 2025, 900 US adults clicked a traditional result on 8 percent of visits with a Google AI summary, versus 15 percent without one [6].
Apple: Eddy Cue testified in May 2025 that “last month, our search volume actually went down” in Safari, and blamed AI tools such as ChatGPT [7].

None of this says Google is dead. People still click, and Forrester says buyers rely on third-party validation to confirm what they see [5][6]. The AI answer sets the first impression, and everything after it is a check on that impression.

I heard the human version on a downtown street, where one kid said to another, “Let me ChatGPT what’s on sale.” When I told that story to John Berryman (Founder, Arcturus Labs), he went further than I would: “I don’t think any website is going to exist in five or ten years.” [8] You can hear the moment on YouTube, see the watch page, or read the write-up. His timeline is a prediction, and you can disagree with it. My claim is narrower and already true: the lookup is happening today, and the engine answering it is not reading your homepage the way you imagine.

Why doesn’t ChatGPT know my company?

For most companies, the engine is not remembering you. It is looking you up.

The research behind retrieval-augmented generation (RAG) began with a limitation: language models hold knowledge in their weights but struggle to access it precisely, give provenance for it, or update it [9]. So engines pair a model with live retrieval. ChatGPT returns answers with links to web sources, using third-party search providers and partner content [10]. With web search enabled, Claude searches when a request depends on information that is current, changing, or outside its training data, including information about specific organizations, and always includes citations [11]. Perplexity searches in real time and numbers its citations [12]. Google’s AI Mode and AI Overviews may use “query fan-out,” issuing multiple related searches across subtopics and data sources [4].

The engine runs several searches about you and your category, then writes from whatever comes back. If that is thin, old, or about someone else, the answer is too. Four things make that likelier than you would expect.

1. Your own site is a small share of what the engine reads. McKinsey puts a brand’s own website at about 5 to 10 percent of the sources AI search references [2]. Ahrefs studied 75,000 brands and found branded web mentions correlated at 0.664 with AI Overview brand visibility, the strongest factor it tested, though correlation is not proof [13]. In Muck Rack’s May 2026 analysis of 1 million+ links cited by ChatGPT, Gemini, and Claude, paid and advertorial content was 0.3 percent of citations and journalism was 27 percent [14]. And the URLs AI engines cite are a different set from Google’s top results: a separate Ahrefs Brand Radar study of 15,000 long-tail queries, reported in Rodrigo Stockebrand’s O’Reilly book on answer engine optimization, found that only 12 percent of URLs cited by ChatGPT, Gemini, and Copilot also appear in Google’s traditional top 10 for the same query, with Perplexity the outlier at roughly 28 percent, which fits its live-web architecture [26].

2. Audience size is not the signal. OtterlyAI analyzed 100 million+ AI citations, and for cited YouTube videos, views, likes, and subscribers each correlated with citation at about zero (r from -0.02 to -0.03), while description length correlated at about 0.31 [15].

3. Small companies start in the lowest tier. A June 2026 preprint tracked 100+ brands and found first-run appearance in relevant AI answers of 73 percent for global household names, 44 percent for mid-market brands, and 11 percent for niche brands [16]. Its author works for Ranqo, which sells AI-visibility tracking, so I treat it as directional. If you are pre-Series B, I would guess you start in that third group.

4. The output is noisy. SparkToro’s Rand Fishkin and Patrick O’Donnell had 600 volunteers run 12 prompts through ChatGPT, Claude, and Google’s AI, 2,961 runs in all. For ChatGPT and Google’s AI, the chance of getting the same brand list twice was under 1 in 100 [17]. Some brands still recurred: Bose, Sony, Sennheiser, and Apple appeared in 55 to 77 percent of 994 headphone responses [17]. The study is not peer reviewed, and O’Donnell’s company sells AI tracking.

A citation is also a pointer, not a guarantee. A Stanford study of four 2023 engines found that, on average, 51.5 percent of sentences were fully supported by their citations and 74.5 percent of citations supported their sentence [18]. A Berkeley paper found users prefer answers with more citations even when the cited content does not support the claim [19]. The engines have changed since, so read those as a warning about method. It is why the gap has a “wrong” shape and not only an “absent” one.

What does a Citation Gap look like in practice?

We ran the check on ourselves. In August 2026 we audited a cohort of past guests who participated in the TwoSetAI Founder Visibility Program (episodes from August 2025 to August 2026), about seven queries each across Google, Google AI Mode, Perplexity, and ChatGPT, in incognito windows [3]. A guest counted as present if any query surfaced them.

95 percent appeared on at least one Google surface, and 86 percent were cited by at least one AI engine [3].
In a recent cohort of 10 program participants, every one appeared in Google’s video results and on at least one AI engine. Google AI Mode and Perplexity each cited 90 percent. When ChatGPT did not cite a YouTube video, it reached guests through third-party authoritative posts (founder profiles, twosetai.com pages) [3].
In an earlier cohort, a few participants had no AI-engine citation, and one was invisible on every surface we checked [3].

No paid spend or coordinated amplification sat behind these results. Same guests, same episodes, four engines, four different answers. The gap is per engine, per query, and it moves. ChatGPT is the engine the written layer earns, and the written pages we build around each episode exist to close it.

The Buyer Bucket: how AI tiers you

A buyer typing your name into ChatGPT does not get a page of blue links. They get an answer. That answer slots you into a tier based on what AI retrieved, not what you actually do. We call the result the Buyer Bucket.

Tier	What the AI says about you	What the buyer concludes	How you got there
Tier 1: Default answer	Specific thesis, named clients, dated proof, cited across multiple pages	”Established, specific, worth a meeting.”	Dense editorial coverage + guest conversations + third-party citations
Tier 2: Mentioned peer	Named alongside 2-3 competitors in a list	”A plausible option; need to compare.”	Some content, incomplete coverage, missing comparison articles
Tier 3: Shallow summary	Generic one-line description, no specifics	”Could be anyone. Skip for now.”	LinkedIn bio + press mentions only; no editorial substance
Tier 4: Invisible	”I don’t have specific information about…”	Did not surface at all.	Nothing citeable exists; the buyer never asked a second question

Your tier is not static. It shifts per query, per engine, per week. The point of the audit above is to find out which tier you are in today for the queries that matter.

The clearest closed gap is Alexander Berger (COO, Bolt.new). In August 2026, in incognito windows, ChatGPT labeled our episode with him “Best overall” and ranked it above interviews with Bolt’s CEO on other shows. Perplexity called it “the best fit” and independently recommended the 21-minute highlight. Google AI Mode cited the interview across multiple sections of his bio [20]. The full stack was live: YouTube full episode and highlight, Medium posts, Reddit posts, and a watch page [20]. You can watch the conversation or read the case study. Those answers were reproducible in August 2026. Yours may differ on your run, so check rather than trust me.

How do you check your own Citation Gap?

Here is the version I would run before any call that matters. It takes about 30 minutes, and How to Audit Your Citation Gap in 30 Minutes walks through it.

Write 20 to 30 prompts a counterparty would type. “What does [your company] do?” “[Your company] vs [competitor].” “Best [category] for [use case].” Aleyda Solis’s checklist starts with 30 to 50 core prompts and a baseline before you change anything [21].
Run each in ChatGPT, Perplexity, Google AI Mode, and Claude, in a private window, several times [17].
Log four things: whether you appeared, whether it was accurate, which tier it put you in, and which URLs it cited.
Report presence rate per engine, not rank. “Appeared in 6 of 20 runs” is a number you can improve.
List the cited URLs that are not yours. Those pages are where engines actually read about you.

Step 5 is a visibility problem in the plainest sense. Hector Hernandez (Co-founder and COO, Speakeasy) was talking about companies scaling AI agents, not brand search, but his line fits: “the first thing that breaks when people scale their use of AI is often not cost… most of the time it’s complete loss of visibility of what’s going on.” [22] I am borrowing the shape of his point, not claiming he said it about search. The watch page has the full conversation. You cannot fix a gap you cannot see.

How do you close the Citation Gap?

Engines write from retrieved pages, so closing the gap means making sure accurate, specific, quotable pages exist where engines look, and that they agree with each other. This is generative engine optimization (GEO), also called answer engine optimization (AEO) or AI-answer SEO. Five moves, in the order I would do them:

Decide what you want AI to say first. Vague positioning produces vague citations. TwoSetAI runs proprietary positioning steps for every founder guest, drawing on classical PMM frameworks and adapted for AI-search citation. An engine can only repeat a claim that exists in specific, quotable form.
Put quotes and sourced numbers on the page. The KDD 2024 paper that introduced GEO tested edits on a 10,000-query benchmark [23]. Cite Sources, Quotation Addition, and Statistics Addition gave the largest gains, up to 40 percent in visibility, while keyword stuffing gave little to none [23]. It is a controlled benchmark, not a promise.
Publish where engines read. In OtterlyAI’s data, 94 percent of cited YouTube videos were long-form and 5.7 percent were Shorts [15]. Google needs only indexing and snippet eligibility, with no special markup [4]. My take: a long conversation with a transcript, chapters, and a written companion page gives an engine more to quote than a landing page. The video pillar makes that case.
Get corroborated off your own site. Given the 5 to 10 percent figure [2] and the mentions correlation [13], start with a dated fact sheet of your official name, description, and claims, then fix conflicting third-party listings. Solis’s checklist has an entity-alignment step for this, then maps which outside sites AI cites in your category [21].
Put your voice in the middle. Quotation Addition was a top performer [23]. In my view, specific first-person stories and numbers are what a competitor cannot copy in an afternoon. I call that the Founder-Voice Moat, covered in the Founder-Voice Moat pillar.

On your own domain, I sort pages into identity, concept explainers, comparisons, frameworks, and persona pages, which I call the 5-Category Architecture (pillar). Google says no special markup is needed [4], so it is not a trick. It makes sure the page an engine needs exists.

Then re-run your prompts after every change. a16z’s Zach Cohen and Seema Amble note that providers are still tuning how models pick sources, so the rules may shift with each model update [24].

Why can’t a monitoring dashboard close the Citation Gap on its own?

Monitoring tools are useful, and I would use one. They measure. Peec AI’s documentation defines Retrieved, Cited, and Citation Rate for the domains behind your tracked prompts [25]. a16z lists Profound, Goodie, Daydream, Ahrefs Brand Radar, and Semrush’s AI toolkit as brand-presence trackers and predicts the strongest companies will go beyond measurement [24]. SparkToro’s authors warn that tools reporting a ranking position in AI are misleading [17].

A dashboard can tell you where you are missing. Some monitoring tools (Profound, Peec AI) now draft pages too. What a dashboard cannot do is own your positioning and your voice. The page it drafts without those two inputs is a template, and a template never becomes a lasting citation. That is the Editorial Layer: human-authored, positioning-anchored content on top of automated monitoring. I go deeper in Why Monitoring Tools Can’t Fix Your AI Visibility.

What the layer does	Monitoring tool (Profound, Peec AI, Otterly)	Editorial Layer (TwoSetAI, in-house content team)
Flags whether you appear in AI answers	Yes	No (not its job)
Measures Retrieved / Cited / Citation Rate per prompt	Yes	No
Shows which URLs engines cite about you	Yes	No
Writes the page that fixes the gap	No	Yes
Anchors content in your positioning + problem statement	No	Yes
Produces founder-voice primary material (not template content)	No	Yes
Builds the hub-and-spoke internal linking structure	No	Yes
Earns the comparison, framework, and persona coverage	No	Yes

Monitoring without an editorial layer tells you where the gap is. The editorial layer is what closes it. Both matter; neither replaces the other.

What should you do this week?

So, why doesn’t AI know your business well enough to recommend you? Because engines assemble answers from a few retrieved pages, most of them not yours, and the pages they find about you are missing, stale, vague, or outweighed by someone else’s. That is the Citation Gap. Closing it is mostly editorial work, and it starts with looking.

This week, run the five-step check with at least 20 prompts. Write the three sentences you want an engine to say about you; if you cannot, that is your real finding. Then list the five non-owned URLs engines cited most, and decide which you can correct, contribute to, or earn.

The lists change run to run [17], so no single screenshot is the whole picture. What you control is whether accurate, specific, quotable pages exist. That is our work at TwoSetAI: we turn one podcast conversation into lasting AI-search presence for AI founders whose buyers check AI before they check the website. When your buyers ask AI, be the answer.

Next in this series: Three Signs Your Brand Has a Citation Gap, The Buyer Bucket, and How ChatGPT, Perplexity, Google AI Mode, and Claude Each Decide What to Cite.

Who is TwoSetAI?

TwoSetAI is one of the top AI podcast and video studios in Silicon Valley, and the team behind this article, hosted by Angelina Yang. The show has featured 100+ AI founders and leaders on how real AI products get built, sold, and trusted. Alongside the podcast, TwoSetAI runs the Founder Visibility Program that turns each conversation into the articles, clips, and pages that make its guests easy to find. Watch full episodes on YouTube at youtube.com/@TwoSetAI, and find every guest and episode at twosetai.com.

Glossary

The Citation Gap (also called: AI-answer gap, brand invisibility in AI search). The space between what AI says about your brand and what your business actually is.
Counterparty Check (also called: pre-meeting AI lookup). The lookup a buyer, investor, or candidate runs on you in an AI assistant before agreeing to talk.
Buyer Bucket (also called: AI brand tier). The tier an AI engine places your brand in, based on what it can cite about you rather than on what your business is.
Founder-Voice Moat (also called: founder-voice advantage). Content in the founder’s own voice that reads like a real person, not a template.
Editorial Layer (also called: GEO editorial layer). Human-authored, positioning-anchored content on top of automated AI-visibility monitoring.
5-Category Architecture (also called: GEO content taxonomy). A content taxonomy of identity, concept, comparison, framework, and persona pages.

About Angelina Yang

Angelina Yang is VP of AI, a fast.ai fellow, and the host of the TwoSetAI podcast, where she has interviewed 100+ AI founders and leaders about how real AI products get built, sold, and trusted. Find her on LinkedIn, talk to her AI clone on Delphi, watch the podcast on YouTube, or browse twosetai.com.

Frequently Asked Questions

What is the Citation Gap in AI search? It is the space between what AI says about your brand and what your business actually is. It can mean AI never mentions you, describes you wrongly, or places you in the wrong tier [1][2].

Why doesn’t ChatGPT know my company? ChatGPT looks you up through web search and writes from the pages it retrieves. If they are thin, outdated, or about someone else, the answer is too. A brand’s own site is only about 5 to 10 percent of the sources AI search references [2][10].

How do I show up in ChatGPT, Perplexity, and Google AI Mode answers? Publish accurate, specific, quotable pages where engines read, and get them corroborated by outside sources. Quotes and sourced statistics lifted visibility by up to 40 percent in a KDD 2024 benchmark [23]. Nothing guarantees a citation [17].

How can I check what AI says about my company? Write 20 to 30 prompts a buyer would type, run each several times in ChatGPT, Perplexity, Google AI Mode, and Claude, and log whether you appeared, whether it was accurate, and which URLs it cited [17][21].

Why does AI give a different answer every time I ask? These systems are probabilistic. In SparkToro’s 2,961 runs, the chance of the same brand list twice was under 1 in 100 [17]. Repeat each prompt and measure how often you appear.

Do monitoring tools like Profound, Otterly, and Peec fix AI visibility? They measure it, showing where you appear and which sources AI cites [25]. They do not write the pages that close the gap, which is the Editorial Layer [24].

How long does it take to show up in AI answers? Nobody publishes a reliable timeline. Google says meeting its technical requirements does not guarantee crawling, indexing, or serving [4]. Re-run your prompts after each change and track presence rate over time.

References

[1] G2 (2025). “How AI Chat is Rewriting B2B Software Buying: Insights From 1,000+ Decision Makers.” G2 Learn Hub, October 6, 2025 (August 2025 survey). https://learn.g2.com/ai-search-surging-for-b2b-buyers

[2] McKinsey & Company (2025). “New front door to the internet: Winning in the age of AI search.” October 2025 (AI Discovery Survey, 1,927 US consumers, August 2025). https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/new-front-door-to-the-internet-winning-in-the-age-of-ai-search

[3] TwoSetAI (2026). Guest search audit, checked August 10 to 25, 2026, incognito, screenshots on file. Internal data; protocol in the Methodology note.

[4] Google (2026). “AI features and your website.” Google Search Central, accessed October 12, 2026. https://developers.google.com/search/docs/appearance/ai-features

[5] Schmitt, C. (2026). “If Buyers Change How They Search, Marketing Must Change How It Shows Up.” Forrester, June 1, 2026. https://www.forrester.com/blogs/if-buyers-change-how-they-search-marketing-must-change-how-it-shows-up/

[6] Chapekis, A. and Lieb, A. (2025). “Google users are less likely to click on links when an AI summary appears in the results.” Pew Research Center, July 22, 2025. https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/

[7] Rogelberg, S. (2025). “Apple’s Eddy Cue: Safari searches are down for first time in 22 years, and AI is to blame.” Fortune, May 8, 2025. https://fortune.com/2025/05/08/apple-eddy-cue-testimony-google-alphabet-safari-ai-search-features

[8] TwoSetAI (2026). John Berryman (Founder, Arcturus Labs) episode, published August 3, 2026; quote at 28:51. https://www.youtube.com/watch?v=BZwIhAuJMhI&t=1731s

[9] Lewis, P. et al. (2020). “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks.” NeurIPS 2020; arXiv:2005.11401. https://arxiv.org/abs/2005.11401

[10] OpenAI (2024). “Introducing ChatGPT search.” October 31, 2024. https://openai.com/index/introducing-chatgpt-search/

[11] Anthropic (2026). “Web search tool.” Claude API documentation, accessed October 12, 2026. https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-search-tool

[12] Perplexity (2026). “How does Perplexity work?” Perplexity Help Center, accessed October 12, 2026. https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work

[13] Linehan, L. (2025). “An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied).” Ahrefs, May 26, 2025. https://ahrefs.com/blog/ai-overview-brand-correlation/

[14] Muck Rack (2026). “What Is AI Reading?” May 2026 edition. https://muckrack.com/blog/what-is-ai-reading-new-insights-may

[15] OtterlyAI (2026). “YouTube AI Citation Study 2026.” Updated March 2, 2026. https://otterly.ai/blog/youtube-ai-citation-study-2026/

[16] Kumar, P. (2026). “Generative Engine Optimization at Scale: Measuring Brand Visibility Across AI Search Engines.” arXiv:2606.20065, June 18, 2026 (Ranqo, a vendor). https://arxiv.org/abs/2606.20065

[17] Fishkin, R. and O’Donnell, P. (2026). “New Research: AIs are highly inconsistent when recommending brands or products.” SparkToro, January 28, 2026. https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/

[18] Liu, N. F., Zhang, T., and Liang, P. (2023). “Evaluating Verifiability in Generative Search Engines.” Stanford; Findings of EMNLP 2023; arXiv:2304.09848. https://arxiv.org/abs/2304.09848

[19] Miroyan, M. et al. (2025). “Search Arena: Analyzing Search-Augmented LLMs.” UC Berkeley; ICLR 2026; arXiv:2506.05334. https://arxiv.org/abs/2506.05334

[20] TwoSetAI (2026). Alexander Berger (COO, Bolt.new) episode and three-engine check, incognito, August 2026. https://www.youtube.com/watch?v=8C3C4Vhknx8

[21] Solis, A. (2026). “The AI Search Optimization Checklist,” updated September 2026. https://www.aleydasolis.com/en/ai-search/ai-search-optimization-checklist/

[22] TwoSetAI (2026). Hector Hernandez (Co-founder and COO, Speakeasy) episode, published October 5, 2026; quote at 22:32. https://www.youtube.com/watch?v=yhOLsoqSSEY&t=1352s

[23] Aggarwal, P. et al. (2024). “GEO: Generative Engine Optimization.” KDD 2024; arXiv:2311.09735. https://arxiv.org/abs/2311.09735

[24] Cohen, Z. and Amble, S. (2025). “How Generative Engine Optimization (GEO) Rewrites the Rules of Search.” Andreessen Horowitz, May 28, 2025. https://a16z.com/geo-over-seo/

[25] Peec AI (2026). “Domains.” Peec AI documentation, accessed October 12, 2026. https://docs.peec.ai/domains

[26] Stockebrand, R. (2026). Answer Engine Optimization. O’Reilly Media, July 2026. https://learning.oreilly.com/library/view/answer-engine-optimization/0642572275549/

Methodology note. Last verified 2026-10-12. Engine behavior comes from each vendor’s own documentation, dated above. Third-party studies are cited with their dates, and several (Ahrefs, Muck Rack, OtterlyAI, Ranqo, Peec) come from companies that sell AI-visibility or PR products, so treat them as directional. The TwoSetAI audit [3] covered a cohort of past guests, not the full catalog: about seven queries per guest (one Google name search, and two each in Google AI Mode, Perplexity, and ChatGPT), run in incognito windows between August 10 and 25, 2026. A guest counted as present on a surface if any query for it surfaced them. The Alexander Berger results [20] were checked in August 2026 and may differ on your run. “Counterparty Check,” “Buyer Bucket,” and the other capitalized terms are TwoSetAI’s own terms, not industry standards. Nothing here guarantees a citation, a ranking, or a business outcome.