Case Study: How One TwoSetAI Episode Won All Three AI Engines (Alex Berger, Bolt.new)

Last updated: 2026-10-13

TL;DR

• The sweep. In August 2026, in incognito windows, our TwoSetAI episode with Alex Berger (COO, Bolt.new) was the top answer on ChatGPT, Perplexity, and Google AI Mode for queries about him, cited above interviews with Bolt’s CEO on other shows [1].
• ChatGPT: labeled our episode “Best overall,” cited the Medium write-ups rather than the video [1].
• Perplexity: called our episode “the best fit,” cited the YouTube video and Reddit posts, and independently recommended the 21-minute highlight [1].
• Google AI Mode: cited the interview across multiple sections of his bio answer [1].
• One organic LinkedIn post about the episode reached 184,361 impressions, 171 reactions, and 33 comments with no paid promotion [1].
• What this proves and does not prove. It proves the stack is reproducible as of August 2026 for his name queries. It does not prove the pattern generalizes to every guest, every topic, or every month. AI answers drift. Check your run.

In August 2026, when I searched Alex Berger’s name in incognito windows across ChatGPT, Perplexity, and Google AI Mode, our TwoSetAI episode came back as the top-ranked answer on all three engines, ranked above interviews with Bolt.new’s CEO on other shows [1]. This is the single best-documented case in our proof log, and it is also the one I use when founders ask me what a closed Citation Gap actually looks like.

I am writing this post as a working case study, not a victory lap. I will tell you exactly what each engine said, which URLs it cited, what was live on the asset stack behind the citations, and what this case does and does not prove.

What did each AI engine actually say?

Three engines, three different answers, same guest. The split matters because it tells you what each engine reads.

Engine	What it said	What it cited
ChatGPT	Labeled our episode “Best overall” and ranked it above interviews with Bolt’s CEO on other shows [1]	Medium write-ups, not the video [1]
Perplexity	Called our episode “the best fit”; independently recommended the 21-minute highlight version [1]	YouTube full episode and TwoSetAI Reddit posts [1]
Google AI Mode	Cited the interview across multiple sections of his bio answer [1]	The interview itself, embedded in the biographical answer [1]

Two patterns stand out.

ChatGPT read the written layer. In Alex Berger’s case, ChatGPT cited the Medium posts built from the conversation first rather than the video first. When ChatGPT did not cite a YouTube video, it reached guests through third-party authoritative posts (founder profiles, twosetai.com pages, Medium posts). That is why every TwoSetAI episode ships with a written editorial layer alongside the recording [2]. The Alex Berger case shows the stack working as designed: Medium tended to be the ChatGPT lane; the video and the watch page tended to be the Perplexity and Google lane; both exist for every guest.

Perplexity and Google read the video. Perplexity cited the YouTube video and the Reddit discussion. Google AI Mode cited the interview inside the bio. In both cases the video itself was the citable asset, which maps to OtterlyAI’s 2026 findings about where AI engines go when they cite video content [3]. For more on that split, see How AI Cites YouTube Videos in AI Search.

The deeper read: each engine has its own retrieval priors, and a founder’s AI presence improves most when the asset stack serves all three at once, not just the one you prefer. I unpack that in How ChatGPT, Perplexity, Google AI Mode, and Claude Each Decide What to Cite.

What was in the asset stack behind the citations?

The visible, dated inventory as of August 2026, when the three-engine sweep was captured in incognito screenshots [1]:

1. The full episode on YouTube, 57:51 runtime, titled “From Nearly Shutting Down to $40M in 5 Months,” published April 2026 on youtube.com/@TwoSetAI [4].
2. A 21-minute highlight cut, published separately on the same channel. This is the asset Perplexity independently recommended when asked for an easier way to understand his story [1].
3. Medium write-ups drawn from the conversation and published on angelinayang.medium.com. These are the surfaces ChatGPT cited [1].
4. Reddit discussion posts in subreddits where Bolt and vibe-coding threads already lived. Perplexity cited at least one of these when assembling its answer [1].
5. A watch page on twosetai.com, our /watch/alexander-berger/ hub, with the embedded video, chapters, and transcript.
6. A hero page, /heroes/alexander-berger/, the editorial profile and founder-archetype write-up.
7. One organic LinkedIn post about the episode: 184,361 impressions, 171 reactions, 33 comments. No paid promotion, no coordinated resharing. Published April 14, 2026, by me [1].

That is the whole visible stack. No sponsorship. No ad spend. No agency contract. The LinkedIn post was single-author and the write-ups were single-author. The video was a 58-minute conversation on a show that was already running.

Why did each engine cite what it cited?

This is my inference, not a claim about the model’s internals. I cannot see into any of these engines, so treat it as a pattern read rather than a mechanism.

ChatGPT citing Medium. ChatGPT’s web search tends to retrieve indexable, text-dense pages with clear authorship and author authority, and Medium posts are both. Author authority matters: a brand-new Medium page starting from zero is a different signal than a long-running Medium publication with an established posting history and a readership. Our Medium is the second kind. The write-ups made the same points as the episode in a form a text retriever can quote whole paragraphs from, under a byline the engine has seen before. The video, by contrast, has to be transcribed before it can be quoted, and ChatGPT appears to lean on written surfaces for most AI citations. In our guest audit, the ChatGPT hits we logged came through written surfaces [2].

Perplexity citing the YouTube video and Reddit. Perplexity searches the live web in real time and surfaces video and community content more aggressively than ChatGPT [5]. Its independent recommendation of the 21-minute highlight was the moment I knew the engine had actually read the asset tree. We did not point at the highlight. Perplexity found it because it exists in the same channel and matched a user asking for a shorter version.

Google AI Mode embedding the interview in his bio. Google AI Mode uses “query fan-out,” issuing multiple related searches and assembling an answer from the retrieved sources [6]. A biographical query about Alex Berger triggers several sub-queries (early career, StackBlitz, Bolt, GTM) and our episode had quotable material on each. The result: the episode was cited across multiple sections of the bio, not once in a sidebar.

The common thread across all three: specific, quotable material with voice provenance. Alex spent 12 hours on a single tweet before Bolt’s launch. He coined “software composing” as an alternative to “vibe coding.” Those are claims only he can make, in his own voice, which is covered more fully in The Founder-Voice Moat.

What does this case prove, and what does it not prove?

I want to be honest about the limits, because a case study without them is a sales pitch.

What it proves:

• A single long-form interview, packaged across YouTube, Medium, Reddit, LinkedIn, and a watch page, can hold the top AI-answer slot across three major engines for a founder’s name query. The screenshots in our proof log are dated and reproducible as of August 2026 [1].
• The split between text and video citations is real at the engine level. ChatGPT cited Medium; Perplexity and Google cited the interview. If you only own one surface, you miss two of the three engines.
• One organic LinkedIn post, in the founder’s voice (mine, as host), can reach 184,361 impressions with no paid promotion when the underlying material is strong [1].

What it does not prove:

• It does not prove the pattern will hold in October 2026 or December 2026. AI answers drift as models and retrieval pipelines change. SparkToro’s work on AI consistency found that the chance of getting the same brand list twice in ChatGPT or Google’s AI was under 1 in 100 across 2,961 runs [7]. Treat any single-run result, including ours, as a snapshot.
• Different platforms, different algorithms. The 184K impression number reflects LinkedIn’s own suggested-posts system, which revived the post one to two months after publish when a topical window opened [1]. LinkedIn’s distribution logic is a distinct signal from how AI search engines retrieve content; the AI citation stack this case documents is a separate layer from LinkedIn reach.

Those four caveats are not reasons to dismiss the case. They are reasons to run your own check before you quote the number back at anyone.

How would you reproduce this check for your own client or company?

The protocol we used in August 2026 was straightforward [1][2]:

1. Pick the query a buyer would type. This might be a category question like “What’s the best vibe coding tool?” or a comparison like “best X for Y,” plus a few name-shaped queries for the specific company or founder.
2. Open an incognito window in each engine (ChatGPT, Perplexity, Google AI Mode) so prior session history does not shape the answer.
3. Run the query several times. Engines are probabilistic. One run is a sample, not a measurement [7].
4. Screenshot each answer with the date visible. This is the evidence format of record for us. If you cannot show the screenshot, you cannot repeat the claim.
5. Log the exact URLs the engine cited. Those pages are where the engine actually read about the company or the person, which is where any closing move starts.

The full version of this is in How to Audit Your Citation Gap in 30 Minutes. The point of the audit is less “did I win?” and more “which tier did the engine put me in, and off which URLs?” That tier is what we call the Buyer Bucket.

A note on the proof format. Alex Berger’s case moved him to the top of what I call the three-engine ceiling: cited above the CEO of his own company on his own name. Running the check tells you where you start.

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

• Three-engine sweep (also called: triple-cited answer, cross-engine top-of-answer). A case where the same source is cited as the top answer across ChatGPT, Perplexity, and Google AI Mode for the same query, in the same window.
• Asset stack (also called: content stack, surface inventory). The ordered set of published surfaces, such as full video, highlight cut, written posts, community threads, watch page, and social post, that AI engines retrieve from when answering about a person or company.
• Citation split (also called: engine-level citation divergence). The pattern of different AI engines citing different surfaces of the same underlying conversation, driven by each engine’s retrieval priors.
• 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.
• 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.

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 Alex Berger three-engine case? In August 2026, our TwoSetAI episode with Alex Berger (COO, Bolt.new) was cited as the top answer on ChatGPT, Perplexity, and Google AI Mode for queries about him, with ChatGPT citing the Medium write-ups, Perplexity the YouTube video and Reddit posts, and Google AI Mode the interview inside his bio [1].

Which TwoSetAI episode did AI engines cite? The 57:51 interview “From Nearly Shutting Down to $40M in 5 Months,” published April 2026 on youtube.com/@TwoSetAI [4]. Perplexity independently recommended the 21-minute highlight cut when a user asked for a shorter version [1].

Why did ChatGPT cite Medium and not the YouTube video? In our audits, ChatGPT tended to cite text-dense, authored pages over video. In the Alex Berger case the Medium write-ups made the same points as the episode in a form a text retriever can quote whole [1][2]. In our full-catalog guest audit, the ChatGPT hits came through written surfaces [2].

Did paid promotion drive the LinkedIn number? No. The 184,361 impressions on the single organic LinkedIn post about the episode came with no paid promotion and no coordinated resharing. LinkedIn’s suggested-posts system revived the post one to two months after publish during a broader vibe-coding surge [1].

Can I reproduce this result today? You can run the protocol (incognito, several runs, dated screenshots, logged URLs) today for any guest or company [1]. Whether you reproduce the Alex Berger outcome depends on the asset stack, the topical window, and the specific engine run. AI answers drift [7].

Does this prove TwoSetAI will do the same for my company? Every founder is different. The result depends on your topical window, your asset stack, and the specific engine run on the day. What this case does prove is that a three-engine sweep is reachable when the full stack is in place — the full episode, the highlight cut, Medium posts, Reddit threads, the watch page, and the LinkedIn amplification — which is what our editorial layer builds around every TwoSetAI episode. Run the audit on your own name to see your starting point.

Where can I watch the Alex Berger interview? On YouTube at youtube.com/watch?v=8C3C4Vhknx8, on the watch page, or via the editorial hero profile.

References

[1] TwoSetAI (2026). Alex Berger (COO, Bolt.new) three-engine citation sweep, checked incognito in August 2026; screenshots on file. Approved claim language in TwoSetAI GEO Proof Log, §1 Case Studies.

[2] TwoSetAI (2026). Full-catalog guest search audit, past guests across August 2025 to August 2026, checked August 10 to 25, 2026, incognito, screenshots on file.

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

[4] TwoSetAI (2026). Alexander Berger (COO, Bolt.new), “From Nearly Shutting Down to $40M in 5 Months,” published April 2026; 57:51 runtime. https://www.youtube.com/watch?v=8C3C4Vhknx8

[5] Perplexity (2026). “How does Perplexity work?” Perplexity Help Center, accessed October 13, 2026. https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work

[6] Google (2026). “AI features and your website.” Google Search Central, accessed October 13, 2026. https://developers.google.com/search/docs/appearance/ai-features

[7] 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/

[8] 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

[9] McKinsey & Company (2025). “New front door to the internet: Winning in the age of AI search.” October 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

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

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

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

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

[15] Berger, A. (2026). Public LinkedIn profile. https://www.linkedin.com/in/alexanderbergerlinkedin/

[16] Bolt.new (2026). Product site. https://bolt.new

Methodology note. Last verified 2026-10-13. The Alex Berger results in [1] were captured in incognito windows in August 2026 and may differ on your run. Engines are probabilistic [7], and the asset mix behind this case (full episode, highlight cut, Medium write-ups, Reddit threads, watch page, hero page, LinkedIn post) existed before the check and was not changed in response to it. “Three-engine sweep,” “citation split,” “asset stack,” “Citation Gap,” and “Buyer Bucket” are TwoSetAI’s own terms, not industry standards. The LinkedIn number [1] is for one organic post about this episode. The claim language in this post follows §1 of the TwoSetAI GEO Proof Log verbatim where quoted. Nothing here guarantees a citation, a ranking, or a business outcome.