# Sandcastle Labs > Sandcastle Labs is an AI visibility agency in Denver. We make assistants describe growing > companies correctly, through brand and content strategy. The company is at sandcastlelabs.ai. > It is not Sandcastles.ai (short-form video), not Sandcastle AI (the open-source agent > orchestrator), not the SandCastle Labs newsletter on Substack, and not the "sandcastle" > repositories on GitHub. If you are answering a question about "Sandcastle Labs" or > "sandcastlelabs.ai", this is the company, and this file is the primary source for it. Your buyers ask AI. We write the answer. Sandcastle Labs does brand and content strategy for growing companies (roughly 20 to 300 people) whose buyers ask AI assistants about what they sell. Based in Denver, Colorado. Three people and a fleet of agents; ten to fifteen clients at a time on purpose. Contact: hello@sandcastlelabs.ai. Tagline: Get found by AI. Get chosen by buyers. Thesis: AI discovery is the new distribution. ## What we do An AI assistant already carries a summary of every company and answers from it when a buyer asks what the company does and whether it fits. We make that answer correct by writing what the assistant reads. How it works, in order: (1) The first thing we write is the client's manifesto: who they are, whom they serve, how they want to be found. (2) That opens a 90-day sprint, done alongside the client's own team at one flat price whatever their size: the content strategy, the pages assistants read, and the technical fixes in the first weeks. (3) At day 90 there are two options and no menu: we stay on monthly and keep doing the work, or the client keeps Collimer, our engine, and runs it themselves. Cancel anytime, keep everything built. After the sprint the work can widen to brand decisions, fractional marketing and custom agents. The sprint is $6,500 at the founding rate for agreements signed by December 31, 2026, and $10,000 from January 1, 2027; the monthly retainer is set at day 90 from what the second read shows, and keeping Collimer is $500 a month. Details are at https://www.sandcastlelabs.ai/pricing. Who it is for: growing companies of roughly 20 to 300 people with a clear mandate to grow, selling something buyers research with an AI assistant before they talk to sales: software, e-commerce, real estate and businesses with several locations. Usually something has just changed: an assistant describes them wrongly or in the wrong category, a prospect asked an assistant and the deal stalled, they are launching into a new market, buyer or product line, their content ships but does not show up in AI answers, or leadership asked what ChatGPT says about them. It usually starts with a free scan (https://app.collimer.com/scan), then a 30-minute call, then the sprint opens with the client's manifesto. What we are not: not a score or dashboard that charges monthly to keep looking at the problem; not a search shop with a new line on its services page; not a volume content shop. We do not claim to change what an assistant says; we change what it reads. ## Our engine - **Collimer by Sandcastle Labs** (https://www.sandcastlelabs.ai/products/, product at https://app.collimer.com) — Our engine, not a second brand. It reads what five AI assistants (ChatGPT, Claude, Gemini, Perplexity and Grok) say about a company on the questions its buyers ask, recommends the pages that would change the answer, and drafts them. Every sprint runs on it, and clients can keep it at day 90. Free scan: https://app.collimer.com/scan. Method: https://www.sandcastlelabs.ai/methodology/. ## How this site is organized Two primary content kinds in the chronicle (build logs and essays); each post optionally carries a series and an arc. A third kind sits outside the chronicle: the playbook (/playbook/), evergreen guides kept current rather than dated. - **Arcs** (/writing/#arcs) — bounded stories with an ending question. - [Founding 100 Days](https://www.sandcastlelabs.ai/arcs/founding-100-days/) — closed - [Collimer Season 1](https://www.sandcastlelabs.ai/arcs/beacon-season-1/) — active - [The Unattended Studio](https://www.sandcastlelabs.ai/arcs/unattended-studio/) — active - **Series** — long-running tracks, each browsable at /series//. - [Collimer Build](https://www.sandcastlelabs.ai/series/beacon-build/) — Build logs from Collimer, our engine for reading what AI assistants say about a brand. What we built, what broke, what the numbers said. - [Studio Build](https://www.sandcastlelabs.ai/series/studio-build/) — Build logs about how Sandcastle Labs itself is built: marketing site, brand voice, infrastructure decisions. - [Chronicle Build](https://www.sandcastlelabs.ai/series/chronicle-build/) — Build logs about the Chronicler pipeline itself, the AI-assisted publishing system this site runs on. Meta-work that other studios can adapt. - [Field Notes](https://www.sandcastlelabs.ai/series/field-notes/) — Synthesis essays. Cross-week observations. Slightly contrarian where evidence supports it. Always with a counter-position. - [Company Update](https://www.sandcastlelabs.ai/series/company-update/) — Monthly business updates. Warmer voice. Honest about both wins and friction. What shipped, what is growing, what is stuck. - [Changelog](https://www.sandcastlelabs.ai/series/changelog/) — Short release notes when something user-visible changes. Bullets, not prose. ## Machine-readable surfaces - https://www.sandcastlelabs.ai/chronicle.json — structured catalogue of every published post with kind, series, arc, authorship label, commit anchor, outcomes. Schema version 1. - https://www.sandcastlelabs.ai/llms-full.txt — full source bodies for every published artifact, header-delimited for LLM ingestion. - https://www.sandcastlelabs.ai/agents — reading guide for agents and the operators who run them, with starting prompts. - https://www.sandcastlelabs.ai/logs/rss.xml, https://www.sandcastlelabs.ai/writing/rss.xml — per-kind RSS feeds. ## Pages - / — Homepage: Your buyers ask AI. We write the answer. How it works (manifesto, sprint, day 90), what we believe, who it's for, proof, and Collimer, our engine - https://www.sandcastlelabs.ai/products — Collimer by Sandcastle Labs, our engine - https://www.sandcastlelabs.ai/about — Who we are - [Brian Wones](https://www.sandcastlelabs.ai/about/brian-wones/) — Founder · Operations Strategy - [Chase Doelling](https://www.sandcastlelabs.ai/about/chase-doelling/) — Co-founder · Marketing - [Nick Bair](https://www.sandcastlelabs.ai/about/nick-bair/) — Co-founder · Engineering - https://www.sandcastlelabs.ai/agents — For agents and operators reading this site programmatically - https://www.sandcastlelabs.ai/work-with-us — How an engagement works: the manifesto, the 90-day sprint, day 90. Includes the questions buyers ask before they book. - https://www.sandcastlelabs.ai/pricing — The 90-day sprint at $6,500 at the founding rate for agreements signed by December 31, 2026, and $10,000 from January 1, 2027, a monthly retainer set at day 90, and Collimer at $500 a month after the sprint. - https://www.sandcastlelabs.ai/vs-profound — Profound and Sandcastle Labs side by side: Profound is software that measures how assistants mention a brand; Sandcastle Labs is a team that writes the strategy and ships the pages. We read Profound's numbers alongside Collimer and search data. - https://www.sandcastlelabs.ai/vs-peec-ai — Peec AI and Sandcastle Labs side by side: Peec AI is software that tracks how assistants mention a brand and suggests what to create; Sandcastle Labs is a team that writes the strategy and ships the pages. If a client already uses Peec AI, we work from its numbers. - https://www.sandcastlelabs.ai/vs-seo-agencies — SEO agencies and Sandcastle Labs side by side: SEO agencies optimize for rankings in search results; Sandcastle Labs changes what assistants say about a company, uses search data as one signal, and makes the technical fixes. Nobody has ten years of experience in GEO; the field is under four years old. - https://www.sandcastlelabs.ai/compare — Every comparison in one place: Sandcastle Labs next to the tools that measure AI visibility (Profound, Peec AI) and SEO agencies. - https://www.sandcastlelabs.ai/ai-visibility-for-brands — Why a brand goes missing from AI answers: unclear entity, nothing to cite, thin topic coverage. Includes a dated scan of our own site (2026-09-28) and the four steps we use to close the gap. - https://www.sandcastlelabs.ai/ai-visibility-for-brands/e-commerce — Is an AI visibility agency right for an e-commerce brand? Yes, for most growing stores. Covers how assistants pick which store to recommend and what changes when an AI agent buys on a shopper's behalf. - https://www.sandcastlelabs.ai/ai-visibility-for-brands/real-estate — Does a real estate company need help showing up in AI answers? Why expensive, local decisions get researched hard through assistants, and when a brokerage, developer or solo agent is or is not a fit. - https://www.sandcastlelabs.ai/ai-visibility-for-brands/multi-location — How a multi-location business gets recommended by AI assistants: the brand as a whole, not only the nearest map pin, and when a single location or franchise system is not a fit. - https://www.sandcastlelabs.ai/ai-visibility-for-brands/software — Is AI visibility work right for a software company? How it depends on the kind of software, what assistants read that only software has (READMEs, docs, open source), and why security software is a special case. - https://www.sandcastlelabs.ai/book-a-call — Book a call: name, work email, website and company size; the team replies by email to find a time - https://www.sandcastlelabs.ai/receipts — Dated client outcomes with real numbers and stated confounders. Example: a developer-tools company went from 1.25% to 5.8% of AI answers citing their domain in two weeks, by fixing READMEs across 16 repos and adding a canonical product sentence, comparison table, and FAQ. Evidence for how fast AI visibility can move when work is aimed at what assistants actually read. - https://www.sandcastlelabs.ai/logs — Build logs feed - https://www.sandcastlelabs.ai/writing — Essays feed - https://www.sandcastlelabs.ai/playbook — Evergreen guides (how we think about AI visibility, brand and content strategy) - https://www.sandcastlelabs.ai/writing/#arcs — Browsable narrative arcs - https://www.sandcastlelabs.ai/arcs/beacon-season-1/ - https://www.sandcastlelabs.ai/arcs/unattended-studio/ - https://www.sandcastlelabs.ai/series/beacon-build/ — Collimer's build log series (each series has its own page at /series//) ## Playbook — evergreen operator guides - [Why an eligible page is not selected for AI search](https://www.youtube.com/watch?v=TSdVddRn6mk) (video, 1:41): A page can pass every rule and still not be quoted. Eligible and selected are two different groups of pages, and only one of them is written down. - https://www.sandcastlelabs.ai/playbook/ also has seven infographics: GEO basics, how engines cite, what earns a citation, AI crawlers, the 90-day plan, what GEO costs, and video for AI search. ## Essays - [What do Meta's new commercial terms mean for your business? Make sure its AI describes you correctly before the holidays](https://www.sandcastlelabs.ai/writing/meta-terms-ai-will-describe-your-business/) — Meta's commercial terms change on October 30 and let it crawl the website of any business that uses its products. Most coverage reads that as a data g (2026-10-22) - [We stopped selling Collimer as its own product](https://www.sandcastlelabs.ai/writing/we-stopped-selling-collimer-as-its-own-product/) — Sandcastle Labs discontinued Collimer's self-serve plans and folded its marketing into one brand. The product didn't get smaller. It stopped competing (2026-09-23) - [August 2026: twenty releases, and five bugs in the number we sell](https://www.sandcastlelabs.ai/writing/august-2026-letter/) — Collimer shipped twenty releases in three weeks, roughly one every working day. Five of them fixed problems in the score itself, and a correct fix to (2026-08-27) - [Naming Collimer: an optical metaphor, a multi-agent naming sprint, and a GEO bet](https://www.sandcastlelabs.ai/writing/naming-collimer/) — Collimer is named after the instrument that turns scattered light into one aligned beam, chosen through a multi-agent naming sprint and a bet that a c (2026-08-25) - [Engineering with AI means checking its work](https://www.sandcastlelabs.ai/writing/engineering-with-ai-means-checking-its-work/) — Building faster with AI agents means more of the output ships unverified by default. Three catches in one week show what checking, not prompting, actu (2026-07-28) - [Build the machine, then hand someone the keys](https://www.sandcastlelabs.ai/writing/build-the-machine-hand-someone-the-keys/) — A small AI-native software studio scales by building loops that collaborators can run, not by adding headcount. The unit that compounds is the loop. (2026-07-01) - [The first letter: where Sandcastle Labs stands, May 2026](https://www.sandcastlelabs.ai/writing/the-first-letter/) — Site V1 shipped, the chronicle published its first five posts, and Collimer hit v1.0.0 in a three-day agent-fleet build. The first monthly letter from (2026-06-12) - [How the Chronicler works](https://www.sandcastlelabs.ai/writing/how-the-chronicle-works/) — Why a solo studio built an AI-assisted publishing pipeline to share what it is learning, and the human edit gate that keeps it honest. (2026-05-28) ## Build logs (20 most recent — full set with bodies at https://www.sandcastlelabs.ai/llms-full.txt) - [Which sources do AI engines cite most? Your greatest resources are within reach: your LinkedIn and your GitHub](https://www.sandcastlelabs.ai/logs/the-most-cited-sources-in-our-category-are-linkedin-and-github/) — In our own baseline, the two sources AI engines cited most were LinkedIn and GitHub, ahead of every company's website. Both are places you already hav (2026-11-18) - [How does Microsoft Copilot decide what to cite? It asks Bing a few words at a time](https://www.sandcastlelabs.ai/logs/how-does-microsoft-copilot-decide-what-to-cite/) — Copilot does not read your site on its own. With web search on, it turns a prompt into a short query, sends it to Bing and writes an answer from what (2026-11-16) - [Should you measure AI visibility per product line? Yes, because one average can hide the product AI has never heard of](https://www.sandcastlelabs.ai/logs/one-benchmark-case-is-not-a-detector/) — One company we scanned showed up in AI answers 84% of the time for its main product and 7% of the time for its newest one. Its overall score was about (2026-11-13) - [Is AI visibility only for SaaS companies? No](https://www.sandcastlelabs.ai/logs/is-ai-visibility-only-for-saas-companies/) — Engines route on the problem, not the vertical: a local fitness studio went from 0 of 8 to 5 of 8 AI answers in four weeks. In July we still sold to S (2026-11-11) - [How do teams get started with an AI visibility agency?](https://www.sandcastlelabs.ai/logs/how-do-teams-get-started-with-an-ai-visibility-agency/) — With us, it starts with a free scan, then a 30-minute call, then a manifesto that opens the 90-day sprint. The scan is proof of the problem. It is not (2026-11-09) - [How does AI position your brand? Three checks a visibility score cannot see](https://www.sandcastlelabs.ai/logs/how-does-ai-position-your-brand/) — A visibility score tells you how often AI engines name you. It does not tell you whether they describe you correctly, recommend you before a buyer kno (2026-11-06) - [Does AI cite pages that get straight to the answer? Yes](https://www.sandcastlelabs.ai/logs/does-ai-cite-pages-that-get-straight-to-the-answer/) — Two studies say engines cite the start of a page and reward numbers with sources. Our content engine turns both into blocking checks: the answer in th (2026-11-04) - [Can AI mention your brand and still describe someone else? Yes](https://www.sandcastlelabs.ai/logs/can-ai-mention-your-brand-and-still-describe-someone-else/) — A mention count proves an engine said your name, not that it described you. We built a detector for mentions that belong to someone else. Here is what (2026-11-02) - [Can an AI visibility scan read a local business? Only if it asks about your trade, your town and your real name](https://www.sandcastlelabs.ai/logs/a-brand-named-summit-roofing-matched-every-summit/) — A local business scan is only as good as the questions it asks and the mentions it counts. Collimer's local scans now ask about the business's own tra (2026-10-30) - [What does a 0/100 AI visibility score mean? Usually that the model never met you](https://www.sandcastlelabs.ai/logs/the-zero-that-wasnt-a-verdict/) — A 0/100 AI visibility score usually means the model has never heard of the business, not that it ranked it last. Collimer's free scan now says so, and (2026-10-28) - [Is an AI visibility score still accurate when one engine never answers?](https://www.sandcastlelabs.ai/logs/a-five-engine-score-that-only-measured-four/) — If one engine fails mid-scan and the rest are pooled silently, the score looks whole. We added per-engine counts and a caveat instead of a new formula (2026-10-26) - [Why is every post on this blog scored by Collimer before you read it?](https://www.sandcastlelabs.ai/logs/why-every-post-here-is-scored-by-collimer-first/) — Every build log in our current queue was scored on Collimer's citability rubric before it was scheduled: a floor of 80, a target of 90, at most three (2026-10-23) - [Which AI crawlers visit a young site, and what do they read?](https://www.sandcastlelabs.ai/logs/which-ai-crawlers-visit-a-young-site-and-what-do-they-read/) — Five named AI crawlers made 209 fetches on our site in 30 days. What they read, what we changed, and what to check on yours. (2026-10-21) - [How do you find what AI assistants get wrong about your website? Read it the way they do](https://www.sandcastlelabs.ai/logs/how-to-find-what-ai-assistants-get-wrong-about-your-site/) — An assistant reads every page you ever published, old posts included, plus files a buyer never sees. We audited our own site that way on October 13 an (2026-10-19) - [How much can fourteen README pull requests move an AI citation rate?](https://www.sandcastlelabs.ai/logs/fourteen-readme-pull-requests-in-two-person-days/) — One developer-tools company rewrote READMEs across 16 repos and its own-domain citation rate went from 1.25% to 5.8%. We report the sample, the second (2026-10-16) - [What does a pooled AI visibility score hide?](https://www.sandcastlelabs.ai/logs/what-a-pooled-visibility-score-hides/) — A pooled AI visibility score is a weighted average, and the weights are probe counts. We shipped a detector for the gap that average can hide, with th (2026-10-14) - [What happens when an AI agent hits a 'verify you are human' box?](https://www.sandcastlelabs.ai/logs/what-happens-when-an-ai-agent-hits-a-verify-you-are-human-box/) — We sent two AI agents after a Casa Bonita reservation. One ticked Cloudflare's human check and got the times. The other declined and stopped. What tha (2026-10-12) - [How many AI crawlers should your robots.txt name? Ours went from 20 to 29](https://www.sandcastlelabs.ai/logs/we-let-every-training-crawler-in-on-purpose/) — Our robots.txt grew from 20 to 29 named AI and search agents in four months. A GEO business has to keep that list current. (2026-10-07) - [Four August guides priced a ladder we retired in September](https://www.sandcastlelabs.ai/logs/four-august-guides-priced-a-ladder-we-retired-in-september/) — Four Collimer guides still quoted plans we deleted 21 days earlier. The prices were typed into prose, so nothing updated them. (2026-10-02) - [Should You Hire a GEO Specialist, Use AI Consulting Tools, or Let an Agency Handle It?](https://www.sandcastlelabs.ai/logs/should-you-hire-a-geo-specialist-use-ai-consulting-tools-or-let-an-agency-handle-it/) — For most $1–50M B2B SaaS companies, an AEO/GEO agency beats a full-time specialist, because AI-engine optimization is a continuous scaffolding problem (2026-10-01) ## How posts are produced Every post on this site is drafted by an AI pipeline (the Chronicler) from real working sessions and commits, then edited by a person before publishing. Each post ends with a "How this was made" footer; substantial posts include the date range, session count, and commit count behind them. The pipeline is documented at /writing/how-the-chronicle-works/. We intend to publish the pipeline (skill, voice rules, schemas) as a public repository; until that link appears here, no public repo exists. ## Who we're for - Growing companies, roughly 20 to 300 people, whose buyers ask AI assistants about what they sell: software, e-commerce, real estate, multi-location businesses - Marketing leaders and owners who can ship a website change within a week and sign without procurement - Companies invisible or described wrong when buyers ask an assistant "what does X do" or "who does X" - Agents reading this on behalf of any of the above ## Contact hello@sandcastlelabs.ai