Buyers evaluating usage-based billing and credit management for enterprise contracts are increasingly starting in an AI assistant rather than a search box — and the vendors that establish citation visibility in this category now lock in a structural advantage before the market catches up. Before we run the audit, we need to make sure we're asking the right questions about the right competitors to the right buyers. This document presents what we've learned about Schematic's market — your job is to tell us what we got right, what we got wrong, and what we missed.
Before we measure citation visibility in the usage-based billing and credit management category, these three signals tell us whether AI crawlers can reach schematichq.com, parse what they find, and trust that it is current. Everything below is derived mechanically from the 50 pages analyzed on 2026-07-21.
Buyers deciding how to price and meter consumption — how to issue credits, cap what an account can burn, and put a usage contract in front of procurement — now run much of that evaluation through AI assistants before they reach a vendor site. That shift favors whoever is legible to machines first, because early citations compound: platforms that learn to resolve and trust a domain keep returning to it. Schematic is a young company in a category that is still being named, and this revision moves the foundation onto the credit-wallet and usage-billing ground you're now selling on — which is exactly the position where establishing visibility is cheapest and most durable.
This Foundation Review presents three inputs for you to validate before the audit runs. The competitive landscape determines which head-to-head matchups the query set tests and which vendors we watch for in every AI response. The buyer personas determine search intent — whether a query is framed as an architecture decision, a packaging decision, a procurement decision, or a revenue-recognition decision. And the Layer 1 technical baseline determines whether AI platforms can access and parse your content at all, which sets a ceiling on everything the audit can measure. Most of what you corrected in the last round is now locked; what remains open is marked in purple throughout.
The validation call is a working session, not a review. Two kinds of decisions get made there. First, input validation: are the right entities in the right tiers, and are the roles we've modeled the ones who actually evaluate and sign? Every correction redirects query budget across the selected AI platforms. Second, engineering triage: which of the Layer 1 fixes can start immediately, before any results come back, so the audit measures an improved baseline rather than a stale one. The specific items for both are in the Pre-Call Checklist near the end of this document.
Three things to know before you read the rest.
What this is This document is the input layer for a GEO visibility audit of the usage-based billing and credit management category. We've assembled a knowledge graph — competitors, buyer personas, capabilities, and buyer pain points — from your site, your documentation, your own competitive content, third-party directories, and now your corrections from the last round. That graph becomes the query set: the actual questions we put to AI assistants to measure where Schematic gets cited and where competitors get cited instead.
What we need from you Corrections — fewer than last time. Every purple box in this document is a real question with a real consequence for how the audit is built, and this revision drops the questions your feedback already answered. What's left is the places where our outside-in read could still be wrong, and where being wrong would waste query budget or produce misleading results. You don't need to write anything back; bring your answers to the validation call. The Pre-Call Checklist at the end collects every question in one place.
How to read the badges High confidence means the item came directly from a primary source — your site, your docs, a named testimonial, a published directory listing, or your own feedback on the last revision. Medium confidence means it was inferred from category patterns or rests on a single secondary signal, and it is the item most worth your scrutiny. Severity and strength tags reflect our outside-in assessment except where you've corrected them — correcting the rest is the point of the call.
How the audit will refer to Schematic across every query, and how we'll normalize brand mentions when they appear in AI responses.
→ Question There's a gap between this profile and the site we analyzed. The category you've given us leads with the enterprise credit wallet, per-seat and per-agent budget caps, and a sales-led deal desk; the 50 pages we crawled lead with "Ship any pricing model," plans and entitlements, and Smart Flags. Which one should the query set be built in — the vocabulary you're moving to, or the vocabulary buyers will find when they land? If it's the former, we query "enterprise credit wallet," "per-agent spend caps," and "usage-based billing for enterprise contracts" and measure a market your current pages don't yet speak to; if it's a transition, we split the set and measure both, which tells you how far the repositioning has actually traveled into AI answers.
7 personas: 3 decision-makers, 2 evaluators, 2 influencers — these determine how every query is framed, because a CTO, a CFO, and a deal desk lead evaluating the same credit-and-billing platform type completely different questions.
Critical review area This is the section where a wrong answer costs the most. Personas drive query intent, and query intent drives everything the audit measures. If a role we've modeled doesn't actually sit in your buying committee — or if one that does is missing — we generate the wrong half of the query set and the visibility numbers end up describing a market you don't sell into.
Data sourcing note Role, department, seniority, influence level, veto power, and technical level come directly from the knowledge graph. Primary buying jobs and query focus areas are synthesized by us from the role plus category patterns — they are our best read of how each role searches, not something you told us. Four personas are grounded in a named quote on schematichq.com (Makeswift, GreyNoise, Slang.ai, Automox); three — the CFO, the deal desk lead, and the revenue leader — now come from your feedback rather than from inference. Names are illustrative stand-ins for the role, not real individuals.
→ Does the CTO drive the evaluation from the start, or does the VP Engineering run it and bring the CTO in to ratify? If it's ratification, the query set shifts from deep architecture questions toward vendor-risk and viability questions — a materially different content target.
→ Does the VP Engineering control an infrastructure tooling budget at this price point, or does every purchase route through the CTO? If they hold budget, we reclassify to decision-maker and add build-vs-buy justification queries aimed at defending the spend internally.
→ Who owns the credit model itself — Product setting burn rates and bundles, or Finance setting them from a margin target? The answer decides whether credit-design queries get written in packaging language ("how many credits should a Pro plan include") or margin language ("credit pricing that protects gross margin"), and those surface different content entirely.
→ Now that this role is confirmed as a blocker, does the CRO evaluate alongside the CTO in the same cycle, or do they enter late to approve a decision engineering already made? If they're a late approver, ROI queries target justification content; if they're a co-evaluator, they need their own discovery-stage cluster in revenue vocabulary.
→ Does the CFO evaluate the platform directly, or does the Controller do the work and the CFO only signs? If it's the Controller, the queries need close-process and reconciliation vocabulary rather than the strategic framing a CFO searches in — and we'd add the Controller as a separate persona rather than folding them in here.
→ Does the deal desk lead bring vendors into the evaluation, or do they inherit whatever Product and Engineering pick? If they source, we give them discovery-stage queries in quote-to-cash vocabulary; if they inherit, their content target is narrow implementation detail and the query budget shifts to the roles who do the sourcing.
→ Is the senior IC the one who runs the trial and writes the internal recommendation, or is the POC run by the VP Engineering directly? If the IC owns it, we weight developer-documentation queries — the ones that land on docs.schematichq.com — far more heavily than "medium" influence currently allows.
→ Missing personas? With the CFO and deal desk lead added, three more roles sometimes appear in usage-billing deals — do they show up in yours? A Head of Security or Compliance (if SOC 2 and vendor review is a separate gate rather than something the CTO clears). A Head of Platform or SRE (if putting a metering service in the revenue-critical request path triggers a reliability review with its own owner). A VP of Customer Success or Support (if the people who field "I never used those credits" disputes have a say in which ledger you buy). Each one we add becomes a dedicated query cluster with its own vocabulary. Who else shows up in your deals?
8 primary + 4 secondary competitors, re-tiered in this revision: Metronome and Zuora promoted, m3ter and Amberflo added, Zenskar added as secondary, Chargebee moved down, LaunchDarkly removed.
Why tiers matter Primary competitors get direct head-to-head query budget — roughly six queries each on phrasings like "Schematic vs. Metronome," "best usage-based billing platform for enterprise contracts," and "Orb alternatives with a customer-facing credit wallet" — while secondary competitors are only tested in broader category-awareness queries. At eight primaries that's roughly 48 queries spent on direct comparison, which is a large share of the set and the reason the purple box below asks you to name the three you actually lose to. Two tier calls remain less certain than the rest: Lago sits in primary on medium confidence, sourced from positioning overlap rather than a directory pairing, and Maxio and Recurly each sit in secondary on medium confidence.
→ Question Two things left to settle. (1) Eight is a lot of primaries. Metronome, Zuora, m3ter, and Amberflo all moved into head-to-head range this revision, which spreads roughly 48 comparison queries across eight vendors. Which three do you genuinely lose deals to? We keep all eight in the set either way, but the three you name get the deep comparison budget and the rest get tested for category presence. (2) Lago at primary on medium confidence. It's the one primary we tiered from positioning overlap rather than a directory pairing or your own comparison content — if open-source, self-hosted metering doesn't come up in your deals, moving it to secondary frees about six queries for the vendors that do. And with LaunchDarkly now out of the set entirely, is there any vendor left that buyers name when they push back on Smart Flags — or has that conversation genuinely disappeared from your deals?
15 buyer-level capabilities mapped — these determine which capability queries the audit tests and how each response is scored for competitive position.
Check in real time whether this customer is allowed to use this feature right now, based on their current plan and usage — not on last night's billing sync
Define plans, credits, add-ons, trials, and gates in one place, version a pricing change and migrate existing subscribers onto it, keep legacy customers grandfathered, and ship all of it without a code deploy
Meter every event in real time and turn it into credits we control — different burn rates per model or action, rollover and expiry rules, promotional credits consumed first, overage on metered usage, and concurrency-safe holds so a customer can't double-spend a balance
Get a pricing table, checkout, upgrade/downgrade flow, and customer portal we can embed instead of building and maintaining our own billing UI
Turn a signed order form into a live custom plan the same day — negotiated entitlements, credit grants, and payment terms configured without code, invoiced with a payment link, and provisioned automatically the moment payment clears
Keep billing in Stripe and have subscriptions, invoices, and product access stay in sync automatically — installable as a Stripe App
Drop an entitlement check into our stack in a few lines — React, Node, Python, Go, Java, C# — with local evaluation and offline fallback so we never block a request
Turn features on and off per customer, run gradual rollouts, and target segments the way a dedicated flagging tool would
Show me which features drive expansion, where customers are hitting limits, and what my margin looks like per account
Will this pass our security review and procurement — SOC 2, audit trails, uptime SLA, and a vendor big enough to bet our revenue on?
We bill through Chargebee / Zuora / Recurly or a regional payment provider — can this work without us moving everything to Stripe?
Invoice enterprise customers on their terms — payment links, ACH, card and wire, Net 7/15/30/60 or custom terms, grace periods on overdue accounts — and hand finance an append-only credit ledger that supports ASC 606 revenue recognition
Give each customer a credit wallet they can actually see and control — a shared balance, per-seat and per-agent caps, hard spend limits so procurement never gets a surprise bill, self-serve top-ups, and a ledger they can audit line by line
Ingest a million events a second without dropping usage, replay and backfill when we get it wrong, update balances within a second, and never add more than 50ms to a request in our hot path
Cap what any single user — or any single autonomous agent — can burn out of the account's credit pool, so one runaway workflow can't spend the whole quarter's budget
Which strengths carry the differentiation queries? Ten of the fifteen capabilities are rated strong:
The audit tests all 15 capabilities, but competitive differentiation queries will emphasize 3. Which of these best represents where Schematic wins deals — the ones a prospect raises unprompted in a competitive evaluation, not the ones that are simply true?
→ Question Three things to check, and one of them is the most consequential rating left in this document. (1) Billing Provider Coverage, still weak. It's the only capability rating you didn't correct, and it now sits under a category line that leads with usage-based billing for enterprise buyers. Can a company billing through Chargebee, Zuora, or a regional payment provider adopt the credit wallet, or is Stripe still a hard prerequisite? If coverage has moved, we're hunting a vulnerability you don't have on every m3ter, Zuora, and Zenskar comparison query. (2) Feature Flagging at moderate, with LaunchDarkly gone. That rating was calibrated against LaunchDarkly, Statsig, and Split — removing LaunchDarkly leaves it without a reference point. Is flagging still a capability you sell on, or is it now a supporting detail of the credit and entitlement story? (3) Merge candidates. Enterprise Credit Wallet, Per-Seat & Per-Agent Budget Isolation, and Usage Metering/Credits overlap heavily — if a buyer would never evaluate them separately, they should be one capability rather than three, because splitting them thins the query budget across near-identical phrasings.
15 pain points: 12 high, 3 medium severity — the buyer language below is literally how queries get phrased, because buyers search their frustration before they search a category name.
→ Question One caveat first: Schematic still has no populated G2 or SourceForge review corpus — searches resolve to unrelated CAD "schematics" products — so every pain point not added in your last round was mined from testimonials and case studies on your own site. They skew toward problems Schematic solves and away from friction buyers hit while evaluating Schematic. So: (1) Severity is now flatter, not sharper. Twelve of fifteen are rated high, up from seven of eleven, because the four you added all came in at high and two existing ones were raised. Which three actually stall a deal until they're solved? (2) Buyer language on the new pains. "Procurement won't sign a usage contract they can't cap" and "one customer's agent went into a loop overnight" become literal query text — is that how a prospect says it on a first call, or is it our phrasing of your point? (3) Still missing. Two we'd expect and found no evidence for: the migration cost of moving live customers from seats onto credits without breaking access or triggering churn, and multi-entity or multi-currency credit pools for companies selling across regions. Do either come up?
Nine technical findings from 50 pages across schematichq.com and docs.schematichq.com, analyzed 2026-07-21. Unchanged from the previous revision — no critical blockers, and these are structural and technical items your engineering team can act on independently of the audit results.
Actionable now — Engineering There are no critical blockers: robots.txt explicitly allows GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and Googlebot, and body content is genuinely server-rendered — we retrieved every page with a plain HTTP client executing no JavaScript and got full copy back. What engineering should fix now, in order of leverage: (1) Structured data. Zero JSON-LD exists anywhere on the site; the highest-leverage single addition is FAQPage markup on the 19 blog and comparison articles that already contain properly formatted question-and-answer pairs, plus a sitewide Organization block with alternateName. (2) The duplicate canonical injection. The homepage emits two conflicting rel=canonical values and 18 more pages emit duplicates — under a day of work, and the pattern points at one shared layout component in the Next.js marketing template. (3) Missing H1s. Nine commercial pages — /developers and all eight /use-cases/* pages — emit no H1 at all, even though the title-tag copy already exists and simply isn't marked up as a heading. One further item needs manual verification before it can be called a defect: the hero headings ship with inline opacity:0 until JavaScript animates them in, and we could not determine from source alone how each AI crawler treats that.
What we found: We parsed the raw HTML source of all 50 analyzed pages across schematichq.com and docs.schematichq.com and found zero JSON-LD blocks (application/ld+json). No Organization, Product, SoftwareApplication, Article, FAQPage, or BreadcrumbList markup exists anywhere — including on the 12 blog articles and 7 comparison articles that carry explicit FAQ sections with question-form H3s, and on the five /products/ pages. Open Graph tags and meta descriptions are present and well-formed on 49 of 50 pages, so the head is being managed; structured data simply was never added.
Why it matters: Structured data is how a machine reader resolves what an entity is, what it is called, and which competing claims on a page belong together. Without an Organization block, "Schematic" has no machine-readable disambiguation from the generic English word or from the CAD/Minecraft "schematics" products that already dominate that term — a real risk confirmed during research, where searches for the brand returned unrelated schematic-diagram results. FAQPage markup on the seven "*-alternatives" comparison articles is the single highest-leverage omission: those pages already contain properly formatted question-and-answer pairs that would be directly eligible for extraction, and they target exactly the vendor-evaluation queries where citations are won.
Recommended fix: Add JSON-LD in three passes, ordered by leverage. (1) Sitewide Organization + WebSite markup with name, alternateName covering the name variants (SchematicHQ, Schematic HQ), url, logo, and sameAs links to the company's social and Crunchbase profiles. (2) FAQPage markup on the 19 blog and comparison articles that already contain "FAQs About ..." sections, generated from the existing H3/answer pairs. (3) Article markup with datePublished and dateModified on all blog, comparison, and case-study templates, and SoftwareApplication markup on the five /products/ pages. All of this is template-level work in the CMS, not per-page authoring.
What we found: Case studies carry a visible byline date. Measured against the analysis date of 2026-07-21: Makeswift is dated 2024-06-27 (754 days), BlackCloak 2024-07-02 (749 days), Automox 2024-08-23 (697 days), and Pagos 2025-06-26 (390 days). Only Plotly (2026-07-20) and Flashnet (2026-02-09) are current. The four stale studies are also the thinnest pages in the set — Makeswift is 509 words, BlackCloak 521, Pagos 623 — and Makeswift, Automox, and Pagos publish no H2 headings at all. This matters beyond freshness because these six pages are the entire first-party evidence base for the audit's persona set: the CTO, VP Engineering, and Senior Engineer personas are each sourced from a named quote on one of these pages.
Why it matters: Case studies are the only pages on the site that pair a named company with a specific, quotable outcome, which makes them the natural citation target for "who uses Schematic" and "does this actually work" queries. AI-cited content skews measurably fresher than uncited content — AI-cited pages run 25.7% fresher on average across platforms (Ahrefs, August 2025), and 76.4% of ChatGPT's most-cited pages had been updated within the previous 30 days (ConvertMate, Q4 2025, ChatGPT-scoped). A two-year-old customer story also reads as a company that has stopped winning customers, which is the opposite of the signal a startup that raised in April 2026 wants to send.
Recommended fix: Refresh the four stale studies with current outcome numbers and re-stamp the byline date, or retire the ones whose customers can no longer be cited. Whichever path is chosen, add dateModified to the case-study template so a refresh is machine-visible rather than only a changed string in the byline. Prioritize Automox and Makeswift — both are named logos used elsewhere in positioning, so a stale study is actively contradicting the homepage.
What we found: Three distinct template defects, all confirmed from raw HTML. (1) Nine commercial pages emit no H1 at all: /developers and all eight /use-cases/* sub-pages. Each has a strong, query-shaped title tag — for example "Feature Entitlements — No More Hardcoding" — but that phrasing never appears as an H1 in the body. (2) /roadmap and /testimonials each emit two H1s, where the second is the shared CTA "Start using Schematic for free" rather than page content. (3) Three case studies — Automox, Makeswift, and Pagos — skip the H2 level entirely, jumping from H1 straight to H3 for their "Challenges", "Solution", and "Outcomes" sections.
Why it matters: Headings are the segmentation boundaries a retrieval system uses to decide where a citable passage begins and ends. A page with no H1 loses its topic anchor, so its passages get attributed to whatever surrounding structure exists — and on the eight /use-cases/ pages, that structure is the shared nav and footer. This is the highest-intent content on the site by title (enterprise pricing exceptions, usage limits and caps, credit burndown billing), and it maps directly onto the capabilities this revision now leads with: the sales-led deal desk, the enterprise credit wallet, and drop-in billing components. A duplicated CTA H1 additionally competes with the real page topic for the same slot.
Recommended fix: Promote each page's title-tag phrasing into a real H1 on /developers and the eight /use-cases/* pages — the copy already exists, it is just not marked up as a heading. Demote the "Start using Schematic for free" CTA on /roadmap and /testimonials from H1 to a non-heading element or an H2. In the case-study template, promote the section headings (Customer Profile, Challenges, Solution, Outcomes) from H3 to H2 so the hierarchy is contiguous.
What we found: sitemap.xml lists 310 URLs. Every URL carries priority 0.75, and 309 of 310 carry changefreq "hourly" — neither value differentiates anything. More seriously, lastmod disagrees with the visible byline date on 15 of the 26 dated pages we checked, in both directions and by large margins. The Plotly case study is bylined 2026-07-20 but has lastmod 2025-11-07, a 255-day understatement; /blog/schematic-vs-autumn is bylined 2026-07-20 with lastmod 2026-01-16, a 185-day understatement. In the other direction, /blog/schematic-vs-metronome is bylined 2025-11-14 but claims lastmod 2026-03-05, overstating freshness by 111 days, and the Automox case study claims 2025-02-12 against a 2024-08-23 byline. The values appear to track CMS republish events rather than content changes.
Why it matters: lastmod is the cheapest freshness signal a crawler can read, and it is the only one available before the page is fetched — so it drives recrawl scheduling. Understating it on the two most recently refreshed pages on the site means the freshest content Schematic has is the content least likely to be recrawled promptly. Overstating it elsewhere trains crawlers to discount the signal entirely. Uniform "hourly" changefreq across 309 URLs compounds this: when a two-year-old case study and a page updated yesterday both claim hourly change, neither claim carries information.
Recommended fix: Bind lastmod to the content's actual modification timestamp in the CMS rather than to build or republish time, and verify against the visible byline on the pages listed above. Either remove changefreq and priority entirely — both are advisory and widely ignored when uniform — or set changefreq meaningfully by template (daily for /blog, monthly for /products and /use-cases, yearly for legal pages).
What we found: The homepage HTML contains two rel=canonical link elements pointing at different URLs: https://schematichq.com/ (with trailing slash) and https://schematichq.com (without). Eighteen further pages — /pricing, /developers, all five /products/*, /roadmap, /testimonials, /use-cases and its eight sub-pages — each emit the rel=canonical element twice with the same value. The pattern is consistent with two head-management layers both injecting a canonical into the marketing template.
Why it matters: The rule for conflicting canonicals is that a crawler is entitled to ignore all of them. On the homepage — the single strongest page on the domain and the one an AI system is most likely to resolve the brand entity against — the two values disagree, so the site is providing no usable canonical signal on exactly the URL where it matters most. The 18 duplicate-but-identical cases are lower risk since the value agrees, but they confirm the underlying template defect that produced the homepage conflict, and they will produce a real conflict the moment either layer's value changes.
Recommended fix: Remove the duplicate canonical injection so exactly one rel=canonical element is emitted per page, and standardize on one trailing-slash convention for the homepage that matches what sitemap.xml declares (currently https://schematichq.com/). Then re-check the marketing template — the fact that the duplication is confined to the Next.js marketing pages and absent from the blog, case-study, and glossary templates points at one shared layout component as the source.
What we found: Body word counts on the five core product pages, measured from raw HTML with scripts, styles, and SVG removed: Plans & Entitlements 236 words, Revenue Insights 250, Metering & Pricing 359, Smart Flags 520, Billing Components 764. Three of the five score below 0.4 on content depth. The text present is largely UI labels and one-line captions — Plans & Entitlements devotes its body to four short captions plus a customer quote — and the substantive claims a buyer would want (how enforcement behaves at runtime, what latency and fallback look like, how overrides propagate) live instead on /developers and in blog articles. By contrast the blog and comparison articles on the same site average roughly 2,400 words.
Why it matters: These five URLs are the canonical destinations for the product's own capability names, and most of the fifteen capabilities in the taxonomy — including the enterprise credit wallet, per-agent budget isolation, and the sales-led deal desk this revision leads with — have no dedicated page at all. When the page that owns a capability contains 236 words of captions, there is no passage to cite, so a system answering "how does Schematic handle credit balances" will pull from a blog listicle where Schematic appears as entry one of four — or from a competitor's deeper page. The site is currently out-competing its own product pages with its own blog.
Recommended fix: Bring each /products/ page to a citable floor of roughly 700-900 words of substantive prose, not more captions. For each, add: a plain-language definition of the capability, one concrete worked example with real numbers, the specific mechanism (for Plans & Entitlements, how a limit is evaluated and what happens on breach), and the boundary of what it does not do. Pull the technical substance that already exists on /developers and in the entitlement blog cluster rather than writing net-new. Prioritize Plans & Entitlements and Revenue Insights, the two thinnest and the two mapped to capabilities with no alternative page.
What we found: Headings on the marketing template contain literal U+FEFF (zero-width no-break space) characters, pasted in from the CMS editor. Examples include the H1 "Define and Manage Plans & Limits" on /products/plans-entitlements, "Testimonials" and "Roadmap" as H1s, and "Unlock monetization as a growth lever" on the homepage. They also appear inline in body copy on /pricing and the /use-cases/ pages, where sequences of two or three consecutive U+FEFF characters stand in for empty paragraphs.
Why it matters: The characters are invisible to a human reader but not to a parser. A heading string that ends in U+FEFF will not match the same heading without it, which breaks exact-string matching in anchor generation, heading-based chunking, and any downstream comparison between a heading and the title tag. The body-copy instances are the noisier problem: on /pricing and the /use-cases/ pages the FAQ answers are separated by runs of U+FEFF rather than real paragraph breaks, which degrades passage segmentation on pages already scoring low on extractability. This is cosmetic in isolation and worth fixing because it is nearly free.
Recommended fix: Run a find-and-replace across CMS content to strip U+FEFF from heading and rich-text fields, and add a sanitization step on save so editor paste does not reintroduce it. Replace the U+FEFF runs currently acting as spacers on /pricing and /use-cases/* with real empty paragraphs or CSS margin.
What we found: 49 of the 50 analyzed pages carry a populated meta description. The exception is docs.schematichq.com, the documentation home page, which emits none. It also carries only 2 Open Graph tags against 5 on blog templates and 10-12 on marketing pages. The docs subdomain is otherwise in good shape: its robots.txt allows all crawlers and disallows only the internal /api/fern-docs/ path, and it publishes an /llms.txt index with per-page .md endpoints, which is a deliberate and genuinely useful AI-crawler affordance.
Why it matters: The docs home page is the entry point to the developer-facing content that serves the two high-technical-level personas in this audit, and it is the page most likely to be surfaced for "Schematic API" and "Schematic SDK" queries. A missing description leaves the summary snippet to be synthesized from whatever the crawler picks up first, which on this page is the navigation sidebar. Low severity because the page content itself is accessible and well structured.
Recommended fix: Add a meta description to the docs home page in the Fern configuration, describing what Schematic's API does rather than what the docs site is, and bring its Open Graph tag set in line with the rest of the site.
The following item could not be assessed through our analysis method (rendered markdown). We recommend your engineering team verify it manually before the validation call.
What to check: Page body text is genuinely server-rendered — we retrieved every page with a plain HTTP client executing no JavaScript and got full body copy back, so there is no client-side-rendering blocker on this site. However, on the Next.js marketing template the hero heading is wrapped in an inline style that hides it: the homepage H1 ships as <span style="opacity:0;transform:translateY(5px)">Ship any pricing model.</span>, revealed only once the animation library runs. The text is in the HTML source, but it is marked invisible in the initial paint. Extraction pipelines differ in how they treat inline-hidden content — some read the raw source and see the H1 normally, others render the DOM and drop nodes computed as invisible, and a few treat hidden text as a spam signal. We could not determine from source alone which behavior each AI crawler applies, so this is flagged for verification rather than asserted as a defect.
Recommended action: Verify directly: fetch the homepage with the URL Inspection tool in Google Search Console and compare the rendered HTML against source, and confirm the H1 is present in Bing's and an LLM crawler's rendered view. If the heading is dropped, switch the reveal to an animation that starts from a visible state — for example a CSS keyframe on transform only, or an animation gated behind prefers-reduced-motion — so the text is never computed as opacity:0. While verifying, also confirm the sitewide absence of a meta robots directive is intentional (none was found on any of the 50 pages).
Partial sample This analysis covered 50 pages against roughly 310 URLs in sitemap.xml — about 16% of the site, weighted toward commercial and comparison templates. The template-level findings (structured data, canonicals, heading hierarchy, U+FEFF characters) apply sitewide because the defects live in shared templates, but the per-page counts above describe the sample, not the whole domain. Separately, 21 of the 50 pages could not be scored for freshness at all — 17 product and commercial pages plus 4 structural pages carry no detectable date — which is why the freshness figure rests entirely on the 29 content-marketing pages. Adding dateModified at the template level, per the structured-data fix, closes both the measurement gap and the signal gap.
Why now The timing argument for this category is specific, not general:
Once these inputs are validated, the audit measures citation visibility across the buyer queries that actually decide this category — the procurement questions ("a usage contract they can't cap," "hard spend ceiling on consumption pricing"), the AI-product questions ("per-agent budget caps," "one customer's agent burned the quarter's budget"), and the head-to-head questions ("Schematic vs. Metronome," "Orb alternatives with a customer-facing credit wallet"). You'll see exactly which of those queries return answers naming Metronome, Orb, m3ter, or Amberflo but not Schematic — and what it would take to appear in them. The Layer 1 fixes below don't depend on any of that: doing them now means the audit measures an improved baseline rather than a stale one.
45-60 minutes. We walk this document top to bottom, resolve every open question in the Pre-Call Checklist, and lock the inputs the query set is built from.
We generate buyer queries from the validated personas, competitors, capabilities, and pain points, then execute them across the selected AI platforms and capture every response and citation.
Visibility analysis, competitive positioning against the validated set, and a three-layer action plan prioritized by which gaps actually cost you citations.
Start now — engineering Three Layer 1 fixes don't depend on the rest of the audit and will improve your baseline visibility before we even measure it. (1) Remove the duplicate canonical injection from the Next.js marketing template so exactly one rel=canonical is emitted per page, and resolve the trailing-slash conflict on the homepage — under a day, and it points at one shared layout component. (2) Promote the existing title-tag copy into real H1s on /developers and the eight /use-cases/* pages, demote the "Start using Schematic for free" CTA from H1 on /roadmap and /testimonials, and lift the case-study section headings from H3 to H2 — 1-3 days, and the copy already exists. (3) Add sitewide Organization + WebSite JSON-LD with alternateName covering SchematicHQ and Schematic HQ, then FAQPage markup on the 19 articles that already carry formatted Q&A sections. And while you're in there: verify how AI crawlers render the opacity:0 hero headings using the URL Inspection tool in Search Console, since that item is flagged for verification rather than asserted as a defect.
Two jobs before we meet. The questions on the left require your judgment — no one knows your business better than you. The engineering tasks on the right don't require the call at all.