Josh Grant

Last week I turned an old guide into working software. One day. Four prompts. Zero code. Built in Lovable. In February I published a 15 section guide to question mining. It went viral. Without even realizing it, I've written a spec to an insanely powerful tool. What question mining is/why it matters: Buyers stopped browsing ten links. They ask an assistant, and the assistant hands back one synthesized answer. There is no page two. If your explanation cannot survive that synthesis, you are absent at the exact moment the decision gets made. Keywords describe topics. Questions expose uncertainty. Uncertainty is where decisions happen. "Headless CMS" is a topic. "Is a headless CMS overkill if my marketing team needs to move fast" is a decision. What the app does: I click one button. Ninety seconds later I'm looking at every theme of uncertainty in a category, mined live from Reddit, Hacker News, Stack Exchange, and dozens of other sources. Clustered by WHY buyers are asking, never by topic. Scored 0 to 100. Every theme backed by verbatim quotes linking to the exact comments. For this CMS example, top cluster: Architecture Selection Risk. Scored 83. Seven receipts, each one a real human in a real thread. Using a tool like this is much bigger than just AEO. I'm using this real time for StackedGTM right now. A real time read on your market's questions feeds everything: → AEO: you know which answers to own before your competitors find them. → Messaging: verbatim buyer language, dated and linked. No more guessing which phrasing to align with. → Sales: the objections are written down, in the buyer's own words, before the call. → Content: every cluster is a brief with proof attached. → Positioning: you see where uncertainty is massing and stake your claim first. One pipeline. Every downstream team fed. Now the part marketers need to hear: Lovable is one of the most powerful tools a marketing team can touch right now. Custom tooling is the thing marketing has never been able to get built. It's valuable to one team and urgent to nobody, so it dies in prioritization every time. Lovable deletes that gap. Database, auth, scraping, hosting, deploys, etc. I never wrote a line of code. I never waited on a sprint. (Disclosure: I'm a #LovablePartner. It's the tool I use in my business more than any other, and this is the same post I would have written without it.)

These are 5 AEO agents we’re running with our top clients right now. I’m finding that most companies’ AEO automation still stops at production. These automate the entire decision loop: Observe → Diagnose → Act → Verify ✳️ Citation Decay Agent Watches citation coverage across the pages and prompts that matter and flags meaningful drops. When coverage falls, it looks at: -which prompts we disappeared from -which URLs took the citation -what changed in those sources The fix might be fresh data, stronger evidence, or one weak section. Then it reruns the prompts and measures recovery. Refresh because the market changed, not because a 90-day clock went off. ✳️ Fanout Gap Agent Take a question like: "What's the best AEO platform for enterprise SaaS?" The answer may depend on security, integrations, pricing, customer proof, international support and comparisons. This agent maps those questions and finds where we're strong, weak, or absent. Sometimes the fix is a new page. Sometimes it's one section in docs. Sometimes the product already does the thing and marketing buried it four clicks deep. Sometimes (usually) nobody credible outside the company supports the claim. ✳️ Citation Displacement Agent First agent catches the drop. This one decides what happens next. It looks at the source replacing us and routes the response: -Competitor → evidence / positioning -Publisher → PR / earned media -Review site → customer marketing -Outdated third party → outreach -Our content → fix it -Product limitation → product A citation loss shouldn't land in the same queue every time. If it only tells you what happened, you built a dashboard. If it figures out why and triggers the next action, you have something useful. ✳️ Brand Truth Agent Checks what ChatGPT, Google and Perplexity say about the company against what's actually true. Profound's new FactCheck is awesome for this. Pricing. Features. Integrations. Security. ICP. Markets. Positioning. Limitations. Then it traces wrong information back to the source. Our site is wrong? Fix it. Publisher outdated? Contact them. No credible source supports the correct answer? Create one. Share of voice doesn't mean much if the model is telling prospects the wrong thing about you. ✳️ Commercial Shortlist Agent Probably my favorite. Run the questions closest to revenue: -"Best X" -"Best X for enterprise" -"X vs Y" -"Alternatives to X" Then pull apart the answer: -Who got recommended? -What evidence supported it? -Why were we included or excluded? Then classify the loss. They have something we don't? Product. We have it but nobody can tell? Positioning. We say it but nobody credible backs it? Authority. The evidence exists but the answer engine isn't finding it? AEO. Those require different actions. This is the AEO system I want running: What changed? Why? What should we do? Did it work? Observe → Diagnose → Act → Verify Then run it again.

Today, StackedGTM is launching a new arm: Forward Deployed Marketing. I believe in this so much I turned down four CMO offers earlier this year to build it. Late-stage VC backed orgs. Strong boards. Real money. I went deep into every process, and each one surfaced some version of the same admission: "We buy a lot of marketing help. The growth never shows up." The help stays outside the building. The work lives inside the building. That gap is where marketing budgets go to die. Forward Deployed Marketing closes it. Our operators embed inside your company. We build in your stack, alongside your team, and ship working systems: live pages, running pipelines, production agents, and the infrastructure behind them. Fixed scope. 90 days. Your team owns and operates everything when we leave. This is how I've always worked. At Webflow, I built a centralized global growth team embedded across the business and owned an incremental ARR number across the whole channel portfolio. A number you couldn't fake with vanity channel metrics. That same structure produced one of the most respected AEO programs in the industry. At Affirm, I ran growth channels, product growth, and messaging and positioning toward one target: truly incremental, profitable revenue. That discipline carried GMV from $8.3B to $20.2B through an IPO. And over the last few months, I've been running dozens of deployments with some of the top AI and tech brands on AEO. In the weeds, onsite and offsite, tied to incremental results. Embed. Build. Ship. Transfer. We deploy across six areas: AI search and AEO, marketing agents, messaging and positioning, growth channels, marketing engineering, and measurement. StackedGTM remains a media company first. The deployments are what make the media valuable. We test, fail, and win inside real companies, then share all of it with our readers. The content gets sharper because the work is real. The work gets better because 21,000+ readers are pressure-testing it (no pressure 😅). Here's my bet: the next decade of marketing services gets built inside the building, or it doesn't get bought at all. Teams will stop paying to be advised and start paying to be armed. Forward deployed marketing is the future. If you want us in yours: stackedgtm.ai/services

The Reddit, Inc. chart is the symptom. The August 8 change is the disease. Here's the sequence: Aug 8: ChatGPT Search rewires how it retrieves. Domain-scoped searches (site:example.com) jump from 0.4% of its background queries to 17%, overnight. Searches per answer nearly double. Aug 14: Reddit's citation share falls from 3.8% to under 1%. Aug 18-20: Peec AI data (shared via Lily Ray) shows the blast radius. Listicle citations down 50%. Comparison pages down 32%. The fan-out terms that declined most: "vs," "comparison," "top," "best." The modifier that surged: "official." Reddit's response: "no meaningful impact" on the business, since AI answers drive little of its traffic. The stock slid about 10% anyway. Today: still on the floor. OpenAI says it doesn't target individual sites (I believe them)...but that's the scary part. Nobody at OpenAI made a decision about Reddit. The retrieval system changed, and one of the most-cited domains on the internet was collateral. And look at that Peec list again. That's two years of pages built specifically to get cited by AI, and they all got hit in the same week. Reddit just happened to be the famous one. Worth knowing: this is the second Reddit citation collapse in a year. Analysts pinned the 2025 one on a Google data change, and it fully recovered. This one might recover too but don't read that as reassurance, read it as the pattern. What actually changed: - Same weeks, same Reddit, opposite outcomes. ChatGPT wiped it in six days while Google barely moved. There's no such thing as your AI citation share. There's one per engine, and they don't move together. - A citation count isn't influence. One study found ChatGPT pulling Reddit pages constantly while crediting them under 2% of the time. The teams that win this track what the models actually read, not just what shows up in the answer. - Content built to be cited just lost to content built to be true. The listicle farms and comparison-page factories took the hit. The boring official page won. Six days from rewire to repricing...this is the cycle time you're planning around.

Here's a wild one...nearly half of the content answer engines cite is less than 13 weeks old. Profound just shipped Citation Decay, which tracks the full life of every cited page: first cited, rise, peak, half-life. Look at the screenshot below. Pages peaking at 3k+ weekly citations, cut in half within 7 to 14 days. Seeing this every day across the portfolio of AEO clients I support. Those who roadmap, build, and coast get tanked in a couple months. Those who run initiatives as always-on active investments are crushing. You never really finish AEO...you operate the channel. Separately: Ravish Agrawal at Gamma asked Profound for this exact feature last week. They shipped it the same day (🤯). Props.