Connor Gillivan

ChatGPT's Thinking models often run more than 10 searches for one response. You type one prompt. It becomes "ecommerce bookkeeping firms," "Shopify bookkeeping pricing," "reviews of X," and a few more. Your page has to win one of those small searches to even be considered. Not the big prompt. The little ones underneath it.

What ChatGPT does with your prompt, in 5 steps: 1. Decides whether it needs to search at all 2 .Splits your prompt into several smaller searches 3. Only considers sources it already trusts 4. Opens and reads the selected pages 5. Writes the answer from a small set of sources Step 3 is the one almost nobody plans for.

10 Marketing Claude Prompts I run every quarter across our portfolio: Save this. Most founders guess at marketing until it quietly stops working. 1/ ICP Message "You're a brand strategist with 15+ years of experience. Here's who buys from us and why [paste]. Tell me exactly who we should be talking to, and the one message that would make them stop scrolling." 2/ Offer Clarity "Here's our current offer [paste]. Rewrite it so a stranger understands what we do, who it's for, and why it's worth it, in one sentence. Then give me 3 sharper versions." 3/ Hook Bank "Here's the topic of my next post [paste]. Write me 20 scroll-stopping opening lines. Mix curiosity, contrarian, and story angles. Flag the 3 you'd bet on." 4/ Content Repurpose "Here's one post that did well [paste]. Turn it into a week of content: a carousel, 3 short posts, an email, and a script. Keep the core idea, change the angle each time." 5/ Landing Page Audit "Here's my landing page copy [paste]. Tell me where a visitor gets confused, where they lose trust, and the 3 changes most likely to lift conversions." 6/ Email Nurture "Write me a 5-email sequence for a lead who downloaded [thing] but hasn't bought. Each email earns the next open. Sell in the last one, not the first." 7/ Objection Content "Here are the 8 reasons people don't buy from us [paste]. Write me one piece of content for each that answers the objection without sounding defensive." 8/ Competitor Teardown "Here are 3 competitors and their messaging [paste]. Tell me what they say well, where they're weak, and the positioning gap we could own." 9/ Channel Focus "Here's where we've been posting and what each channel returns [paste]. Tell me which one deserves 80% of our effort next quarter and which to drop." 10/ Campaign Debrief "Here's the campaign we just ran and the numbers [paste]. Tell me what worked, what to kill, and the one thing worth repeating next time." The prompts are only half of it. Marketing doesn't reward the loudest brand. It rewards the one that says a true thing clearly, over and over, until the right people believe it. AI can write the copy. It can't tell you what you actually stand for. Which prompt are you running first? ♻️ Repost if you know a founder winging their marketing.

In 2023 I started typing prompts into ChatGPT the way our buyers would. "Best bookkeeping service for a Shopify store doing $2 million a year." Then I opened the sources. Half of them weren't in Google's top ten. The model was reading a different internet than the one I'd spent years optimizing for.

Most people think ChatGPT searches once and summarizes. It searches many times, opens a handful of pages, and writes from the ones it trusts. Your job is to be in that handful. Most businesses have never checked whether they are.

She was dumpster diving for food at 15. Shoplifting books at 17. Bounced between ten schools. Dropped out of community college. Never held a job longer than six months. At 22, she started selling vintage clothes on eBay from her bedroom. By 28, that bedroom operation was doing $100 million a year with 200+ employees. Inc. named it the fastest growing retailer in America. Then it went bankrupt. Sophia Amoruso's story isn't a clean arc of inspiration. It's a case study in what happens when marketing instinct outpaces operations. Here's what stands out to me. 1. Quality is distribution. She had no marketing budget. Just a camera, thrift store finds and an eBay account. But every listing looked editorial. Styled photos. Descriptions that told a story. You could spot a Nasty Gal listing without seeing the store name. 2. She knew her customer because she was the customer. On MySpace she targeted followers of magazines like Nylon. Women 16 to 26 in LA and New York. She replied to every comment and message. Revenue went from $223,000 in 2008 to $23 million by 2011. 3. She moved from rented land to owned land. In 2008, eBay banned her over hyperlinks in customer feedback. She launched NastyGal.com instead. Inventory sold out on day one. Bootstrapped on profits, no outside money until it was already huge. On eBay she was a tenant. On her own site she was the landlord. 4. Marketing without operations is a time bomb. Underneath the press, the bestseller and the Forbes cover, things were breaking. She struggled to delegate. The systems never caught up. Overspending, inventory problems, people leaving, a cash flow crisis. She stepped down as CEO in 2015. Nasty Gal filed for Chapter 11 in 2016. Boohoo bought it for $20 million. 5. Her biggest failure became her credibility. She called the bankruptcy her PhD in entrepreneurship. Then she built Girlboss Media, launched Business Class and started Trust Fund, a venture fund backed by Marc Andreessen and David Sacks. She talked about what went wrong openly. That honesty became the positioning. This is the part I keep coming back to. The flashy stuff gets all the attention. Content, branding, social. But the boring stuff keeps the business alive when growth hits. SOPs. Reporting. Clear owners for every job. You can't scale what you haven't systematized. Marketing brings the customers in. Operations is what keeps them. What would happen to your business if growth doubled tomorrow? Would the systems hold?

My 15-Step Perfect AI SEO Writing Checklist: (How I make sure my AI SEO written content is TOP notch) Most AI content sounds like AI content. That's not a model problem. It's an input problem. Generic in, generic out. So every page our system writes goes through 15 steps. Before a human ever reads it. Phase 1: The Brand Brain (before a word is written) 1. Voice dump ↳ 20,000 words of me actually talking. Podcasts, newsletters, calls. 2. Proof library ↳ Real case studies and real numbers, mapped by topic. 3. Opinions list ↳ The positions I'll defend. Especially the annoying ones. 4. Banned list ↳ Every word and phrase I'd never say. Gone forever. 5. Customer language ↳ Pulled from sales calls, support tickets, reviews and Reddit. Phase 2: The Draft (written in my voice) 6. Search + AI research ↳ Every ranking page, and every source AI is citing. 7. Intent-first outline ↳ Structure built from what already wins. 8. Answer in 100 words ↳ The target question gets answered right at the top. 9. My voice, not AI's ↳ Our writing agent, Ember, drafts straight from the brand brain. 10. Real authority ↳ The actual case study. Not "we're experts in this space." Phase 3: The QA (before it ships) 11. Slop scan ↳ No tricolons, no hedging, no tidy mic-drop endings. Trip several and it's rewritten, not edited. 12. Truth pass ↳ Every stat traced to a source. No source, it comes out. 13. Search pass ↳ Headings, internal links, schema and an FAQ block. 14. Human pass ↳ Does the hook earn the next line? Is there a position someone could disagree with? 15. Conversion + sign-off ↳ The CTA matches what the article argued. A human approves every page. The checklist isn't the magic. The loop is. Hundreds of my edits went back into the system. The same mistake doesn't ship twice. Which of these 15 does your AI content skip? ♻️ Repost to help someone fix their AI content. P.S. Want to see this run on your own site? Join the 10 free articles Beta Program. Email [email protected] to see if you qualify.

If your traffic chart is dropping and you're about to cut SEO, read today's issue first. The number you're watching was never the one that mattered. Just posted.

New issue is live: We lost thousands of visitors a month to AI Overviews. Our leads didn't move. The flipped funnel, the EcomBalance pages that prove it, and the 4 keyword types still worth your money. Free to read. Link below.

He was employee #30 at Facebook. Reported directly to Zuckerberg. Had 0.1% equity. Nine months later, he was fired. That equity would be worth over a billion dollars today. Then he joined Mint(.)com as employee #4. Took them from zero to a million users in six months. Was promised 1% of the company. Then he left. Mint sold to Intuit for $170 million. Next came a payments company for social games. Facebook sued him. He shut it down and spent the next few years in his aunt's basement and on his mom's couch. Then he started AppSumo with $50. Today it does over $100 million a year. Noah Kagan's story isn't a clean success narrative. It's a case study in how expensive lessons build great operators. A few things stand out to me. 1. He documented everything before he had an audience. He'd been blogging on OkDork since 2000. Not for followers. Writing was how he thought. A decade later, that library became credibility nobody could copy. 2. He built demand before the product existed. Mint had 100,000 email subscribers before launch. No ads. Just content, email and a real relationship with the TechCrunch writer covering personal finance. I've built every business around this one. 3. He validated with money, not compliments. AppSumo's first deal was Imgur Pro. He bought licenses at $7, sold them at a profit and proved the model in a weekend. His rule: if three people pay you in 48 hours, you have a business. If they don't, you don't. 4. The email list became the business. One Dropbox for Life giveaway added 250,000 subscribers. He ran one every month for six months. Every deal after that had buyers waiting. 5. He learned to separate himself from the company. Getting fired from Facebook sent him into a year of depression. His self-worth was tied to the badge. Most people skip this part of his story. I'd tell any founder to sit with it. The results: eight businesses past $1M in revenue. A New York Times bestseller in Million Dollar Weekend. He even tested 50 covers with Facebook ads before picking one. He was inside two of the most successful companies of a generation and didn't profit from either. That could have been the whole story. Instead it became the playbook. Build the audience first. Get paid before you build. The most useful experience you'll get is usually the one that didn't go to plan. The work is pulling out the lesson and using it on the next thing. What's one lesson from your biggest failure you've never shared?

Tomorrow I'm starting a new series on where the money is in search now. AI ate the top of the funnel. The bottom is where the leads are. Authority is how you get there. First issue tomorrow: the flipped funnel.

Tomorrow: two pages from the same website that explain everything that happened to SEO this year. One lost thousands of visitors a month to AI Overviews. The other keeps sending us clients. What the difference is and the 4 keyword types still worth writing for. Saturday morning.

Steal my AI SEO Backlink Process (to rank on ChatGPT & AI): ChatGPT doesn't pick favorites. It repeats its sources. Ask it "best X for Y" and look under the answer. There's a list of pages it pulled from. Get on those pages and you get into the answer. Here's the exact process: Step 1: Find the prompts ↳ List 20-50 questions your buyers actually ask AI. ↳ "Best X for Y." "X vs Y." "Alternatives to X." ↳ The best ones come straight from your sales calls. Step 2: Run them and open the sources ↳ Ask ChatGPT, Google's AI Overviews and Perplexity. ↳ Open the source list on every single answer. ↳ Run each prompt twice. Answers shift. Step 3: Log every source ↳ Put each cited URL in a Google Sheet. ↳ Tag the type: listicle, directory, review site, forum, YouTube. ↳ Start with the pages cited across multiple prompts. Step 4: Pitch the listicles ↳ Email the author of every "Best X" post that's cited. ↳ Ask to be added, and give them a reason: real data, a case study, an expert quote. ↳ Offer to help them refresh an outdated post. Step 5: Claim the directories ↳ Build out your G2, Clutch and review profiles completely. ↳ Ask happy clients for reviews. ↳ AI reads these as proof from people who aren't you. Step 6: Show up in the threads ↳ Answer Reddit and forum threads with real, useful help. ↳ Always say who you are. ↳ Help first. Mention your business later, if it fits. Then re-run your prompts every week and track who mentions you. This is still link building. It's just aimed at the pages AI already trusts. Most AI citations come from third-party sites, not yours. So your best AI SEO move might happen on someone else's site. Which source type are you going after first? ♻️ Repost to help someone get their brand into AI answers. P.S. Want content built to get cited too? Join the 10 free articles Beta Program. Email [email protected] to see if you qualify.