Filmcuts AI Visibility Audit

The single most important sentence in this audit: ChatGPT, Perplexity, and Claude cannot currently read filmcuts.io at all. When their crawlers identify themselves honestly, the server returns HTTP 403 Forbidden. Five out of seven major AI crawlers we tested were blocked. Only Google AI Overviews (via Google-Extended) and Apple Intelligence (via Applebot-Extended) get a successful response. That alone explains why Filmcuts is absent from every AI-search citation we tested for queries the brand should own ("shot on film stock footage", "authentic analog footage for brand campaigns", "Super 8 footage subscription").

The second story is the content layer underneath. Even when a crawler does get through, what it sees is a JavaScript shell. The homepage, about page, resources page, and pricing page all serve "Processing" as the H1, no real H2 structure, and zero JSON-LD schema. Product pages are better — they have real titles and descriptions — but no VideoObject, Product, or Organization schema anywhere on the site. And the third story is authority: the founders have a serious commercial reel (Aman, Amway, Park Hyatt) that is completely invisible to AI because it lives in a third-party interview rather than on filmcuts.io itself.

The three things that matter

• You are invisible to ChatGPT, Perplexity, and Claude. Bot blocking at the server or CDN layer returns 403 to GPTBot, PerplexityBot, ClaudeBot, OAI-SearchBot, and ChatGPT-User. This isn't a robots.txt issue (robots.txt is permissive). It's an infrastructure rule sitting above robots.txt. Fixing it is a one-day job.
• The site reads as empty to bots that do get in. H1 is "Processing" on every key page except clip detail pages. Zero JSON-LD schema. No definition block. No FAQ. The brand has no extractable answer for "what is Filmcuts" anywhere in server-rendered HTML.
• The authority is real but hidden. Austin's prior client list — Aman, Amway, Park Hyatt — is the kind of signal AI weighs heavily for citation. It's currently only on one third-party interview, not on filmcuts.io. Surfacing it on-site, in extractable form, is a one-hour fix that disproportionately moves citation eligibility.

The goal was to answer one practical question: when an AI assistant tries to read filmcuts.io, what does it actually see, and is that the version of the brand Austin and Fin would want it to cite? We bypassed JavaScript entirely (which is what AI crawlers do) and read the raw HTML each crawler is served — then compared what the brand sells against what AI is currently telling buyers when they ask category questions.

This is the single largest finding in this audit. We sent eight HTTP requests to https://filmcuts.io/, each identifying as a different crawler. Five came back blocked.

The pattern is clean: anything that announces itself as a major LLM crawler gets a 25-byte error response. Anything that looks like a Google or Apple property gets the full page. robots.txt is not the problem. It explicitly allows everything (we read the file directly). The block lives one layer above robots.txt — most likely a Cloudflare bot-management rule or a Next.js middleware rule keyed on user agent.

The practical consequence: when a buyer asks Perplexity "where can I license shot-on-film footage for a brand campaign", Perplexity tries to fetch Filmcuts, gets a 403, and routes the citation elsewhere — to Filmsupply, to Artgrid, to Filmpac, to Stockfilm. Every one of those competitors returns 200 to the same request. We tested.

In the last twelve months a quiet platform shift began: large language models and AI shopping agents started looking for a specific set of files that brands publish to make themselves machine-readable. Shopify auto-publishes most of them. Custom Next.js stacks (like Filmcuts) have to ship them deliberately. Today, Filmcuts ships none.

Two of these matter immediately: /llms.txt and /agents.md. They take an hour each to write and they're the canonical place an AI looks for "tell me about this brand in 200 words". The next two — UCP and MCP — are the early-mover infrastructure for agent-driven commerce. A buyer's AI agent will eventually want to license a clip directly through the MCP endpoint, not by sending the buyer to a browser. Stores set up for it today will be the ones agents transact with first.

Filmcuts has a specific advantage here that most stock-footage libraries lack: the catalog is structured (clips, packs, artists, licenses) and the licensing model is two-tier (Editorial $49, Agency $999) — clean enough to expose as an MCP tool with very little adaptation. It's a strategic moat that's available to seize for a quarter of effort.

This was the most surprising single thing we found. AI crawlers anchor extraction to heading tags — they use the H1 to understand page topic and H2s to understand section structure. When we read the raw HTML that Filmcuts ships to a crawler, here's what we got on the homepage:

The same pattern appears on /about-us and /resources: H1 is literally the loading-skeleton string "Processing", and no real H2s exist in the server response. Everything else hydrates after JavaScript runs in a browser. AI crawlers don't run JavaScript. They read "Processing" and move on.

What to add (suggested)

<h1>Real Shot-on-Film Stock Footage, Licensed for Brand Use</h1>

<p>Filmcuts is a stock-footage library of authentic Super 8,
16 mm, and 35 mm footage, captured by working filmmakers and
scanned at up to 4K ProRes. Founded in 2022 by Austin Divine
and Fin Matson. Used by directors and brand teams who need
analog texture without the cost of an original shoot.</p>

Section H2s should describe the section directly: Browse the Library, How Filmcuts Differs From Digital Stock, Used By (the brand logo strip), Pricing, What Editors Say, Originals: New Films We Funded This Quarter. This should ship as server-rendered HTML, not client-only.

Filmcuts is built on Next.js, which is capable of server-side rendering. The current build is not using it for the pages that matter most. Here's the pattern across the site:

The good news inside this: clip detail pages are server-rendered properly. Their H1 is the real clip name and their meta description is well-written and evocative. The catalog itself is technically discoverable.

The bad news: the pages that define what Filmcuts is as a brand — homepage, about, pricing, resources — are exactly the pages that AI crawlers can't read. The catalog is visible but the story behind it isn't.

Schema (JSON-LD) tells AI what kind of content a page is. It's how a model knows "this is a product, with this price, and this name, and these reviews" versus "this is general marketing text". We checked every key page on filmcuts.io. There are zero JSON-LD blocks. None.

What's missing — and what each unlocks

For a stock-footage library, the single most expensive schema to leave out is VideoObject. Every clip on the site should have one. With it, the catalog becomes addressable by AI as individual licensable assets. Without it, the catalog is just decoration around an unstructured page.

AI systems don't just cite the most relevant page, they cite the most credible one. The credibility signals are domain age, founder identity, press coverage, third-party reviews, and proprietary data. Filmcuts has more of these than is visible to AI today.

This is the most underused asset in the audit. Austin spent close to a decade directing video for international brands including Aman, Amway, and Park Hyatt before launching Filmcuts. That is exactly the kind of credential AI looks for when deciding whether to cite a source on filmmaking topics. Today, it lives on a single third-party site. If it lived in a Person + sameAs JSON-LD block on a real /about page, plus a one-paragraph bio with named clients above the fold, the citation eligibility lift would be measurable inside a month.

The og:url leak

Every page on filmcuts.io declares its canonical Open Graph URL as https://filmcuts.com — not the live .io domain. We tested filmcuts.com directly: the connection times out completely (HTTP 000, no response). Any AI system that uses the og:url tag to verify a canonical source for Filmcuts content will see an unreachable domain and downweight the page. Any human who hears "Filmcuts" and types .com never lands on the brand. This is a single line of metadata to change.

Alternate-TLD coverage

Name collision risk

Filmcuts shares a name with at least three other entities AI can confuse it with:

• FilmCut — an Android video editor app on Google Play. Different category, different intent, but the search query "filmcut" returns this first for any user typing without the trailing 's'.
• Filmcuts Studio (filmcutsstudio.com) — a separate production company. Same name, different industry. AI conflation risk on "filmcuts pricing" or "filmcuts review" queries.
• FilmCut (Plotter Depot) — a heat-transfer vinyl product. Unrelated, but pollutes any branded search.

The mitigation isn't engagement with any of them. It's making Filmcuts's first-person identity unmissable: a structured Organization schema block, a real H1 that includes the word "footage" or "library", named founders with Person schema, and a Wikidata entity creation. The clearer the entity, the harder it is for AI to conflate.

You said you weren't sure what queries to target. Here's how to think about it. AI search visibility splits into four query types, ordered by how fast Filmcuts can realistically win each.

Order of operations: branded first (one to two weeks of on-site work, fully controllable). Educational and category in parallel (one to two quarters). Listicle as the long-haul play. The Originals quarterly-funded-filmmaker program is the easiest unique angle to lead with — there is no other shot-on-film library globally running an artist-funded model like that, which means AI has no competing source for "how Filmcuts supports filmmakers" type stories. Use it.

Ordered by impact divided by effort. The first four items together close most of the "AI can't read us" problem in under a week of engineering time. The middle group is the content layer. The final group is the slower authority play that runs in parallel over a quarter.

• Unblock GPTBot, PerplexityBot, ClaudeBot, OAI-SearchBot, and ChatGPT-User at the CDN or middleware layer. The single most important change in this audit. Verify after the fix with curl -A "GPTBot/1.0" https://filmcuts.io/ returning HTTP 200. 2–4 hrs, infra
• Move homepage, /about-us, /pricing-plans, and /resources to server-side rendering. Add a real H1 and brand definition block to each. Verification test: the words "Super 8" and "$999" should appear in raw curl output of those pages. 1 day, eng
• Fix og:url everywhere. Change the canonical from https://filmcuts.com to https://filmcuts.io across the site (it's almost certainly a single environment variable or layout component). 10 min, eng
• Add explicit Allow blocks to robots.txt for each AI bot. Defensive hygiene — implicit allow becomes explicit, so future infrastructure changes can't quietly break crawl access. 10 min, eng
• Ship /llms.txt and /agents.md. A 200-word brand summary and an agent-instruction file. These are the canonical files AI looks for when asked to "tell me about Filmcuts". 1 hr
• Add Organization, WebSite, and Person JSON-LD to the global layout. Founders' names with sameAs links to Instagram and LinkedIn. Office location. Founding year. This is the entity-level disambiguation block. 1 hr, eng
• Add VideoObject schema to every clip page. Title, description, thumbnail, contentURL, uploadDate, duration. The single most valuable structural change for a stock-footage library. Templatable across the catalog. 2–4 hrs, eng
• Ship a structured FAQ page with FAQPage schema. 10–12 Q&As covering: what is shot-on-film, Super 8 vs 16mm vs 35mm, what license do I need for a commercial, what's the difference between Editorial and Agency, how is Filmcuts different from Artgrid / Filmsupply, are the clips truly captured on film or just emulated. Each Q&A wrapped in JSON-LD. 3 hrs
• Write a real /about-us page with named founders and prior client list. Austin's commercial reel (Aman, Amway, Park Hyatt) is currently invisible to AI. Surface it in extractable form on-site. Add Person schema. 2 hrs
• Run a citation baseline. Test 20 target queries on Perplexity, ChatGPT (web search on), and Google AI Overviews. Log who's cited today. This is the "before" measurement that we re-run in 4 weeks to see what moved. 2 hrs
• Editorial expansion: 6 deeper pieces in /resources. Topics: the shot-on-film vs digital comparison, how to use analog texture in brand work, a filmmaker interview series (one per Originals filmmaker), a behind-the-scenes of the scanning pipeline. Named author, attributed data, structured HowTo sections. 6 weeks of cadence
• Submit Filmcuts and Austin Divine to Wikidata. Establishes the entity so AI assistants can disambiguate against FilmCut (Android app) and Filmcuts Studio. Free, takes about 30 minutes once. 30 min
• Set up third-party review surface. Trustpilot, G2 (or Capterra), and a ProductHunt launch. Email the last 6 months of customers to seed initial reviews. 2 hrs setup, then ongoing
• PR play for listicle inclusion. Pitch Wrapbook, Studiobinder, No Film School, Cinema5D, and Hypebeast for inclusion in their "best stock footage sites" roundups. Each inclusion converts directly to AI citations within weeks. Quarter-long timeline; the slowest lever in the audit but the highest ceiling. Ongoing, 1 quarter to first results
• Ship the AI agent commerce layer. Universal Commerce Protocol manifest and an MCP endpoint that lets agents query the catalog and initiate licenses. This is the long-bet play. If it ships in the next quarter, Filmcuts is the only shot-on-film library agents can transact with directly. 2 weeks, eng — strategic

We've now run this audit across a few dozen brands. Filmcuts's starting position is unusual: the brand is in a category that's about to be very interesting to AI search (analog texture is the explicit trend Filmsupply called out in its 2026 commercial filmmaking report), the founders have a serious commercial reel, the catalog has a unique angle (real film, not digital), and the Originals program is the kind of editorial story AI hunts for because it can't get it anywhere else.

What's holding it back is almost entirely technical. The crawlers are blocked. The HTML they would read if they got in is a JavaScript shell. The schema doesn't exist. The authority signals are real but off-site. Each of these has a clean fix, and the order they're fixed in matters: you can't measure citation lift from content work if the crawlers can't reach the content. Section 11 is sequenced to fix that root cause first.

This audit is a snapshot of one moment in time. AI search changes fast: schema models shift, new crawlers appear, citations rotate every few weeks, and competitors who are also reading their analytics will not stand still. A static document can tell you what to fix. It can't tell you whether the fixes are working, which is where the measurement layer comes in.

What Filmcuts can ship this week (no tooling needed)

• Unblock the five AI bots at the CDN or middleware layer. This is the prerequisite. Nothing else in the audit moves the needle until this is done.
• Fix the og:url leak and move the homepage and pricing pages to SSR. Together with the bot unblock, this is the minimum viable surface for AI to read the brand.
• Ship /llms.txt, /agents.md, and an Organization + Person JSON-LD block. The minimum agent surface.

The measurement layer: Node AI's AI Visibility platform

For the prompt and scan side, we use our own AI Visibility platform. It's the same tool we run for our other clients, and it's how we know whether each fix is moving the needle in AI search rather than just in our intuition.

The platform replaces the "test 5 queries on 4 platforms every Monday" workflow that most brands try to run by hand and abandon within a month. It's the same operating layer we use for our other clients to say exactly which model invented which fact about the brand, and which fixes shifted recognition by which percent.