--- title: "The AI Shortlist: What Eight Founders Learned When They Checked What AI Says About Their Companies" date: 2026-08-10T19:03:38Z modified: 2026-08-13T15:29:26Z permalink: "https://spotlightonstartups.com/the-ai-shortlist-what-eight-founders-learned-when-they-checked-what-ai-says-about-their-companies/" type: post status: publish excerpt: "" wpid: 7961 categories: - Expert Insights tags: - Expert Insights - AEO - AI Citation Accuracy - AI Search - Answer Engine Optimization - founder stories featured_image: "https://spotlightonstartups.com/wp-content/uploads/2026/08/AEO-Cover-image-scaled.png" featured_image_alt: AI search accuracy for founders author: Gregg Kell timestamp: 2026-08-13T15:29:26Z --- By **[Gregg Kell, Editor](https://www.linkedin.com/in/greggkell/)**, **[AI First Earned Media Platform](https://spotlightonstartups.com/wp-content/uploads/wp-mfa-exports/page/home.md)** Most founders have never typed their own company name into ChatGPT, Gemini, or Perplexity and actually read the answer closely. They assume that if the facts are on the website, the facts will make it into the answer. That assumption is where AI search accuracy for founders quietly breaks down. Spotlight on Startups asked eight founders and executives to run that check on their own companies through a project called The AI Shortlist, and each one found something worth fixing. The problem is rarely a factual error a company can point to and dispute. It is something harder to catch: a category stated correctly but a business model flattened, a rebrand the model has not caught up to, a correction that reached the website but never reached the file AI systems actually read, a founder’s name present in one answer and missing from another for reasons that have nothing to do with how established the company is. The solution is not a tool subscription or a one time audit. It is the habit these eight founders stumbled into on their own, simply asking an AI system what it knows about their company and reading the answer the way a prospective customer would. Their experiences, read together, sort into four distinct failure modes any founder can check for this week. **At a glance, here is what each founder found:** | Founder | Company | AI Tool(s) Checked | What They Found | | --- | --- | --- | --- | | **[Joel Goldstein](https://www.linkedin.com/in/joelgoldstein/)** | **[Mr. Checkout Distributors](https://mrcheckoutdistributors.com/)** | ChatGPT | Correct industry, wrong business model, described as a single distributor instead of a 1,100 member network | | **[Dr. Shaun Digan](https://www.linkedin.com/in/shaundigan)** | **[Startup.Ready.](https://startupready.ai)** | ChatGPT, Gemini, Perplexity | Consistent positioning across all three, but one tool still cited an old domain after a migration | | **[Alix Gallardo](https://www.linkedin.com/in/agnamihira/)** | **[Invent](https://www.useinvent.com/)** | Perplexity, ChatGPT | Accurate and current, but sourced almost entirely from the company’s own channels | | **[Mary Ann Romans](https://www.linkedin.com/in/maryannromanscontentmarketing/)** | **[Gray Decision Intelligence](https://www.graydi.us/)** | Not specified | A single Reddit post carried more influence on brand perception than years of owned content | | **[Will Warren](https://www.linkedin.com/in/will-warren11/)** | **[Admita](https://admita.io/)** | ChatGPT | Accurate current positioning mixed with an outdated company name and a discontinued domain three months after a rebrand | | **[Daniel Wilson Kemp](https://www.linkedin.com/company/lifted-holdings)** | **[Lifted Holdings](https://liftedholdings.com/)** | AI assistant feed, llms.txt | A retracted claim kept appearing in the machine readable feed after the human facing correction was already live | | **[Anthony Guerriero](https://www.linkedin.com/in/anthonyguerriero/)** | **[The Leveraged Years](https://www.theleveragedyears.com/)** | Perplexity | Older company was named and attributed correctly, newer company was described accurately with no founder identified | | **[Laura Patten](https://www.linkedin.com/in/laurapatten/)** | **[Thrive on a Dime](https://thriveonadime.io/)** | Gemini | Founder name appeared unprompted in an AI generated answer despite being the newest and smallest company in the group | #### Why AI Gets a Company’s Category Right and Its Business Model Wrong AI answer engines often describe a company’s industry correctly while missing what the company actually does, because they tend to favor whichever page states a claim simply over whichever page states it comprehensively. **Joel Goldstein**, President of **[Mr. Checkout Distributors](https://mrcheckoutdistributors.com/)**, sent one of the clearest diagnoses in this batch. His company connects CPG brands to the right regional and category distributors across a network of more than 1,100 wholesalers, restocking over 50,000 independent grocery, convenience, and pharmacy retailers. When he checked how AI tools described the business, they got the industry right and the model wrong. “I asked ChatGPT what my own company does and it described us as a distributor. We are a network of 1,100 of them,” Goldstein said. “The answer was not malicious, it was assembled from our oldest pages. AI search does not read your best content, it reads your clearest content. We had spent twenty years optimizing for a search engine that asks for keywords and none optimizing for one that asks for an answer.” Goldstein’s diagnosis matches how researchers describe AI retrieval systems. [Perplexity’s own documentation on how it selects sources](https://www.perplexity.ai) explains that its retrieval layer scores pages on relevance, freshness, and structural clarity rather than on how comprehensive a company considers its own best material. A page can rank well for search engines and still get skipped by an answer engine if the actual claim inside it is buried or difficult for the model to extract cleanly. Goldstein’s oldest pages were technically live and indexed. They answered the wrong question clearly, while his newest material sat structured for keywords rather than for a direct answer. His fix was not a tool subscription. He rewrote his site to answer the questions brands actually ask him on the phone, with the network structure stated plainly near the top of the page instead of implied further down. #### Why the Same Company Can Get Different Answers From Different AI Tools Goldstein’s fix worked because he checked one tool and rewrote for clarity. The next founder’s experience shows what happens when a company gets that clarity right and still watches it drift, simply because different AI tools do not update on the same schedule. **Dr. Shaun Digan**, Founder and CEO of **[Startup.Ready.](https://startupready.ai)**, ran a different kind of test. Instead of checking one AI tool, he checked three, ChatGPT, Gemini, and Perplexity, and compared what came back. “I checked how ChatGPT, Gemini, and Perplexity describe Startup.Ready, and the substance surprised me,” Digan said. “All three had our six pillar framework, and two echoed our decision intelligence positioning almost word for word. Gemini had our current site. ChatGPT was still pointing people to our old domain from a recent migration. Perplexity’s version was thinner and a little off.” Digan had already done the hard part correctly. His positioning, a structured six pillar startup readiness framework distinct from a business plan or a generic checklist, was consistent across every model he checked. What he found instead was a freshness problem. Perplexity in particular weighs recency heavily for evolving topics, and [independent analysis of its citation behavior](https://www.google.com/search?q=perplexity+citation+freshness+ranking+study) has found that content published or meaningfully updated within the last thirty days receives a measurable citation boost, a window that compresses further for fast changing subjects. A company can build one clear, ownable category and still watch it drift out of sync across different AI tools without ongoing maintenance. Getting the positioning right once is necessary. It is not sufficient on its own. #### Why Owned Content Alone Does Not Earn AI’s Trust Digan’s drift problem was about timing. The next two founders surfaced a different question entirely: even when an AI answer is accurate and current, whose voice is it actually built on. **Alix Gallardo**, CEO of **[Invent](https://www.useinvent.com/)**, checked Perplexity for the first time in months and found a genuinely good result, with a catch worth noticing. “It used to describe Invent in one paragraph. Now it’s a bullet list of our top features, accurate, with more than half the sources being ours: our website, blog, YouTube, Instagram, X,” Gallardo said. “Every follow up question was architecture and data. Integrations, security, compliance. It nailed each one. ChatGPT, concise and accurate too.” That is a strong result on the surface. It is also a company validating itself almost entirely through its own voice. Perplexity’s citation pipeline draws from roughly ten candidate pages per query and narrows that to three or four sources in the final answer, weighing relevance, freshness, and page structure as the dominant signals. Owned content can clear that bar. What it cannot do on its own is provide the independent, third party confirmation that tends to separate a company AI trusts from a company AI merely describes accurately. **Mary Ann Romans**, AVP Marketing at **[Gray Decision Intelligence](https://www.graydi.us/)**, described the same dynamic from the other direction, through an unrelated but telling anecdote about her own company’s reputation online. “What surprised me was how one Reddit post that lumped us in with several other companies could sometimes carry more weight than all of the authoritative content we’d spent years creating ourselves,” Romans said. “It was a great reminder that AI doesn’t just learn who you are from your website. It learns from the whole conversation happening around your brand, whether you’re part of that conversation or not.” Owned content earns a company the right to be found. It rarely earns the company the right to be trusted on its own. That distinction shows up again and again in how modern answer engines are built, weighing authority signals from outside a brand’s own domain alongside whatever the company says about itself. #### Why a Rebrand Can Take Months to Reach AI Search Gallardo and Romans found a source problem. The next founder found a timestamp problem, and in his industry, a stale timestamp carries more weight than most. **Will Warren**, CEO of **[Admita](https://admita.io)**, a marketing agency for addiction treatment and behavioral health centers, ran his check after a rebrand three months earlier and found a mix of accurate and stale information sitting side by side. “What surprised me was how confidently it mixed that accurate information with outdated and invented details,” Warren said. “We rebranded three months ago, but ChatGPT still referenced our former company name and linked to the old domain, even though that business no longer exists and the old domain now redirects to our current domain, Admita.io. It also claimed we offer branding and content marketing, which are not services we currently provide.” For most companies, that kind of drift is a minor annoyance. For Admita, working in behavioral health, the stakes run higher. A prospective client in crisis searching for accurate information about available care is a different kind of AI answer to get slightly wrong than a company simply outgrowing an old logo. #### Why a Website Correction Does Not Always Reach AI Systems Warren’s problem was a stale rebrand sitting in public view. The next founder found something quieter and more unsettling: a correction that worked everywhere a human would look and nowhere a machine would. **Daniel Wilson Kemp**, Founder and Chief Executive Officer of **[Lifted Holdings](https://liftedholdings.com)**, a payments and software company based in Nashville, uncovered a version of this problem most founders have never thought to check for. “We corrected the advisory itself within ninety minutes, and I assumed that was the end of it,” Kemp said. “It wasn’t. The file we publish for AI assistants and our machine readable feed both kept asserting a claim we had just retracted, because the correction never propagated past the human readable page. Your website isn’t your only description of you anymore, and the machine readable copies rot silently. Nobody notices until someone quotes one back at you.” Kemp’s company publishes an llms.txt file, a small text file that gives AI systems a curated map of a site’s most important content, separate from the pages a human visitor browses directly. The format was first proposed in September 2024 by [Jeremy Howard of Answer.AI](https://llmstxt.org) as a companion to the decades old robots.txt standard, built specifically for AI systems rather than traditional search crawlers. What Kemp discovered is a structural risk almost nobody is checking for yet: a human readable correction and a machine readable file can describe two different versions of the same fact, and the second one is invisible unless someone thinks to look. > **Quick facts on how AI answer engines select and cite sources** > > Perplexity narrows roughly ten candidate pages down to three or four sources per answer, weighing relevance, freshness, and structural clarity as the dominant ranking signals. > > Content published or meaningfully updated within the last thirty days receives a measurable citation boost for fast changing topics, according to independent analysis of Perplexity’s selection behavior. > > The llms.txt standard was first proposed in September 2024 by Jeremy Howard of Answer.AI, built specifically to give AI systems a curated content map separate from what a human visitor browses. #### Why AI Attaches a Founder’s Name to Some Companies and Not Others Kemp’s file kept repeating a fact nobody meant to say anymore. The last two founders raise a different question: not whether the facts are current, but whether AI attaches a name to them at all. **Anthony Guerriero** runs two companies and, without quite intending to, ran a controlled comparison between them. He is a co founder of **[Manhattan Miami Real Estate](https://manhattanmiami.com)**, a dual market luxury brokerage he has operated since the mid 2000s, and the founder of **[The Leveraged Years](https://theleveragedyears.com)**, an AI training company for senior professionals about a year old. “I asked Perplexity to describe both this week,” Guerriero said. “The brokerage came back right. It named me and my partner, got the Manhattan and Miami focus and the foreign buyer work, and cited our own site, LinkedIn, and the BBB listing. The training company was the more instructive miss. It described the business accurately, then said the results did not identify a named founder. Accurate, and attached to nobody.” Both of Guerriero’s businesses depend on trust in a named individual, a licensed broker and CPA on one side, a personal AI mentor for professionals explicitly skeptical of hype on the other. His side by side test shows that accurate and attributed to a real person are two separate outcomes in AI search. **Laura Patten**, founder of **[Thrive on a Dime](https://thriveonadime.com)**, a coaching marketplace connecting clients with affordable holistic coaches, described the opposite outcome. “I was telling the HVAC person working on my mini split about my startup,” Patten said. “He looked it up on his phone, and the Gemini results included, at the end, founded by Laura Patten. He said, wow, you’re one impressive lady. Out of all the things Gemini could have said, it included my name. I was very surprised. To me, that’s not an important detail, but I guess to AI, it is.” Patten’s business is early and small next to Guerriero’s, yet Gemini volunteered her name anyway. Guerriero’s newer company, further along and better resourced, did not get the same treatment. Company age and size do not appear to predict which way this goes. That gap, present across four of the eight founders featured here, may be the single most useful thing to check for in your own AI footprint this week: search your own company name, and notice whether your name shows up next to it, or not. [Independent analysis of how answer engines weigh authority](https://www.google.com/search?q=how+ai+answer+engines+weigh+founder+authority) has found that consistency across multiple independent sources, not just a founder’s own bio page, tends to settle the question for AI systems trying to decide who actually leads a company. #### The Pattern Underneath All Eight Line these eight up and a company’s size, age, or funding stage explains almost none of it. A thirty-five-year-old distribution network and a coaching marketplace less than a year old each ran into a version of the same blind spot. A company can be doing everything right by traditional marketing standards and still be describing itself to AI in a way that quietly shrinks the business, drops a name, or repeats a claim the company itself already walked back. None of these eight founders hired an agency to find this. They asked a question and read the answer carefully. That is the entire method, and it is available to any founder willing to spend ten minutes checking. The harder part is fixing what the check turns up, and staying current as the models themselves keep changing what they reward. Spotlight on Startups has covered this problem from other angles as well, including how [earned media functions differently from paid and owned media in AI search results](https://spotlightonstartups.com/wp-content/uploads/wp-mfa-exports/page/home.md) and what a [founder can control about AI citation signals](https://spotlightonstartups.com/wp-content/uploads/wp-mfa-exports/page/home.md). A founder profile built around clear, attributable, third party reporting, the kind AI systems already lean on for authority signals, does more for a company’s AI footprint than another round of website copy ever will. That is the premise behind every spotlight Spotlight on Startups publishes, and it is worth checking your own name against before you assume your story is already being told correctly. --- **Frequently Asked Questions** **What is AI search accuracy, and why does it matter for founders?** AI search accuracy is how correctly tools like ChatGPT, Gemini, and Perplexity describe a company when someone asks about it directly. It matters because a growing share of buyers form a first impression of a business through an AI generated answer rather than a website visit, and that answer can be outdated, incomplete, or missing entirely even when nothing is factually false. **Why does AI describe a company’s category correctly but get the business model wrong?** This happens because answer engines tend to pull from whichever pages read as the clearest and most extractable answer, not necessarily the most current or comprehensive one. Older pages written for search engine keywords can outrank newer, more accurate pages simply because they state a claim more plainly. **Can the same company be described consistently by one AI tool and inconsistently by another?** Yes, this is common. Different AI tools draw from different source sets and weigh freshness differently, so a company can appear correctly in one tool and show outdated details, like an old domain, in another, especially in the weeks following a rebrand or migration. **Does citing a company’s own website count as a strong AI answer?** Citing a company’s own website counts toward a strong AI answer but is not the whole picture. Independent third party sources tend to be what separates an answer AI states confidently from one it merely repeats. **What is an llms.txt file, and why does it matter if a company issues a correction?** An llms.txt file gives AI systems a curated map of a site’s key content, separate from what a human visitor browses. If a company corrects a claim on its main website but does not update this file or a related machine readable feed, an AI system can keep repeating the retracted claim even though the human facing page is accurate. **Does a founder’s name need to appear online for AI to attribute a company to them?** A founder’s name does not always need to appear for attribution, but consistency across independent sources, not just a founder’s own bio page, appears to help AI systems reliably attach a name to a business. **What is the simplest first step for checking a company’s AI footprint?** The simplest first step is searching the company name directly in ChatGPT, Gemini, and Perplexity, and comparing what each one says against what is actually true today.