The Toronto Startup Website Audit 2026

The short version:

• We scored 50 funded Toronto startups on a machine-verified 20-point AEO Readiness Test. Access is wide open: 90% allow AI crawlers and 93% render without JavaScript. The structure behind that access is missing. Only 5% deploy FAQPage schema, 2% publish machine-readable pricing, and the median company captures just 30% of the points that take deliberate AEO work.
• 95% of the AEO points Toronto startups earn come from framework defaults and off-page profiles, not from anything they built for AI search. The median AEO score is 10.75 out of 20.
• The median composite score is 59 out of 100. Toronto startups are strong on positioning (median 8 of 10) and weak on content (median 5 of 10) and on deliberate AEO.
• Eighteen of 43 companies (42%) scored 7.5 or higher on the content dimension, a sign of consistent publishing. The ones that clear that bar, Tailscale, StackAdapt, League, tend to dominate their categories.
• One in ten target companies could not be audited because the site would not load. When we re-fetched in June, the set of failing sites had shifted, which points to intermittent bot-blocking rather than a stable outage.
• The six top scorers (League 89, Clearco 84, StackAdapt and Tailscale tied at 83, Cohere 81, Relay 80) share a repeatable pattern: clear positioning, consistent content, and trust signals built on purpose. Only three companies score 7.5 of 10 or higher on AEO: Tofu, Relay, and Clearco.

Why we did this

Nobody owns the Toronto startup marketing-data beat. BetaKit covers funding rounds. The Globe covers exits. MaRS publishes ecosystem reports. None of them audit the actual marketing health of the startups they cover.

Stratezik can. We spent May and June 2026 auditing 50 funded Toronto and GTA startups across six dimensions: positioning clarity, AEO readiness, technical health, paid media presence, content output, and trust signals. Every score traces back to a saved screenshot or a fetched-page artifact, and every finding is anchored in the dataset we are publishing alongside this report.

This is the first year. In 2027 we will re-audit the same 50 companies and report the year-over-year movement. By 2028, what changed becomes the story.

Methodology

Sample definition: Toronto and GTA-headquartered startups that publicly announced a funding round between January 2024 and April 2026, with a live public-facing website on the date of audit (May and June 2026). We excluded stealth-mode companies with no website, companies acquired or wound down before the audit date, and companies headquartered outside the GTA even if Toronto-incubated.

Sample size: 50 target companies identified from BetaKit funding announcements (January 2024 to April 2026), cross-referenced against the Crunchbase Toronto/GTA filter, the Communitech member directory, and the MaRS portfolio. Of the 50, 44 returned enough data for a composite score, which requires at least two verifiable dimensions. Four companies, spanning sports tech, event tech, consumer lending, and edtech, could not be scored on any dimension in either fetch window. Two more returned only a single verifiable dimension, including an e-commerce resale company that was unreachable in the May window but reachable during the June AEO re-fetch, which is itself a data point on accessibility consistency. Several others returned partial data and are scored only on the dimensions we could verify. Unscored and partially scored companies are identified in the dataset, available on request, and are not named in this report.

Audit period: May 15, 2026 to June 30, 2026.

Scoring rubric: Each company scored on six dimensions, each worth 0 to 10 points. The composite score is the simple sum (max 60), normalised to 0 to 100 for the headline number. A single human reviewer scored each company to keep scoring consistent, and every score traces to a row in the published dataset.

The six dimensions

1. Positioning clarity (0 to 10): Can a first-time visitor say who this company serves and what makes it different within 10 seconds of landing on the homepage? Scored by the founder-eye test.

2. AEO readiness, the 20-Point AEO Readiness Test (scored out of 20, normalised to 10): Eight criteria worth 2.5 points each. The original four: LocalBusiness or Organization schema present in JSON-LD; FAQPage schema with real Q&A; answer-first formatting, where the first visible text states what the company does and for whom; and a robots.txt that allows GPTBot, CCBot, Google-Extended, PerplexityBot, and ClaudeBot (an absent robots.txt counts as allowing all). Four criteria were added in June 2026, before publication: an llms.txt file present and validly formatted (H1 plus markdown links, with partial credit for a malformed file); JavaScript accessibility, meaning core page content is present in the raw HTML without JavaScript execution, since AI crawlers including GPTBot, ClaudeBot, and PerplexityBot generally do not run JavaScript, so a client-side-only site is blank to them; clear pricing with Product, Service, or Offer schema carrying price fields on the homepage or pricing page, with partial credit for a visible pricing page without schema (this accommodates contact-sales companies that declare their pricing model); and off-page entity alignment, meaning a findable LinkedIn company page, a Crunchbase profile, and domain consistency across them.

Six of the eight criteria are machine-verified by fetching and parsing each site's HTML, robots.txt, llms.txt, and JSON-LD, with no model judgement involved. Entity alignment uses search-index presence checks. Answer-first formatting is the one judgement call, scored against a fixed rubric on the raw homepage text.

A note on llms.txt: it is an emerging and contested standard. Adoption sits near 10% of domains globally per SE Ranking's 300,000-domain study, Google has said on the record that it does not use the file, and no citation lift has been measured. We score it anyway, for two reasons. The file's real consumers today are AI coding and research agents that fetch it directly, and adoption itself is a trackable year-over-year signal. We report it as a leading indicator, not a ranking factor.

The dimension is scored out of 20, then halved to 10 so it carries the same composite weight as every other dimension. We report both numbers.

3. Technical health (0 to 10): Core Web Vitals on the mobile homepage, LCP under 2.5s, INP under 200ms, CLS under 0.1 (4 points); mobile-friendly per Lighthouse (2 points); HTTPS with a valid certificate and no mixed content (2 points); sitemap submitted and indexable (2 points). Note: this dimension is excluded from the composite this iteration because of PageSpeed Insights API rate-limiting (HTTP 429) across the full sample.

4. Paid media presence (0 to 10): Active in the Meta Ad Library (2 points), apparent Google Ads via similar-tool signals or visible UTM patterns (2 points), LinkedIn ads for B2B-targeted companies (2 points), retargeting pixels present (2 points), conversion event evidence (2 points).

5. Content and founder-led signal (0 to 10): Four or more blog posts published in the last 90 days (2.5 points), founder-led LinkedIn content visible (2.5 points), a newsletter signup with a real lead magnet (2.5 points), press or third-party mentions linkable from the site (2.5 points).

6. Trust and authority signals (0 to 10): Real photos of real people (2 points), named case studies or testimonials with logos (2 points), press mentions or genuine awards (2 points), live, identifiable customer reviews on G2, Capterra, Reddit, or Google (2 points), founders identifiable as Person entities with LinkedIn linked from the About page (2 points).

Unverifiable scores: If a dimension could not be cleanly scored for a given company (for example, paid presence that cannot be confirmed either way), we marked it unverifiable in the dataset and left that company out of that dimension's average. We did not invent midpoint scores.

Data sources: BetaKit funding announcements (January 2024 to April 2026), the Crunchbase Toronto/GTA filter, the Communitech member directory, and the MaRS portfolio list. Source registries: betakit.com/tag/funding, crunchbase.com/hub/toronto-startups, communitech.ca/members, marsdd.com/portfolio.

We wrote this methodology before drafting any findings. One change was made before publication: in June 2026 the AEO dimension grew from four criteria to the eight-criterion 20-Point AEO Readiness Test described above, and every company was re-scored on all eight by automated check. No other dimension changed. The rubric constrains the findings rather than the other way around, and it is frozen as of publication for year-over-year comparison.

Top findings

Finding 1: Toronto startups earn their AEO points by accident

We ran every auditable company through the 20-Point AEO Readiness Test, eight machine-verified criteria covering what an AI system needs to find, read, and cite a website. The results split cleanly into two groups of criteria, and that split is the story.

On the criteria that come free with modern infrastructure, Toronto startups score close to perfect. Thirty-seven of 41 companies with a checkable robots.txt (90%) allow all five major AI crawlers: GPTBot, CCBot, Google-Extended, PerplexityBot, and ClaudeBot. Thirty-nine of 42 (93%) serve their core content in raw HTML, readable without JavaScript. Every one of the 50 has a findable LinkedIn or Crunchbase presence. Across the three criteria you get by default, the average company captures 95% of available points.

On the criteria that need someone to build for AI search on purpose, the ecosystem falls away. Two of 42 companies (5%) deploy FAQPage schema. One company in the whole sample, Relay, publishes Product or Offer schema with machine-readable price fields. Twenty-two of 42 (52%) lack basic Organization schema. Only 12 of 40 (30%) pass the answer-first test, where the first visible sentence states what the company does and for whom. Across the five deliberate criteria, the average company captures 29% of available points.

The median total is 10.75 out of 20. No company scored zero, and no company scored above 17. Nobody is fully invisible, and nobody is fully ready.

One genuine surprise: 14 of 42 companies (33%) publish an llms.txt file, roughly triple the 10% global adoption rate. llms.txt is an emerging, contested standard, and Google has said it does not use the file, but that adoption rate suggests Toronto startups are hearing about AI search. They are doing the easy file and skipping the structural work.

When Perplexity answers "best Canadian fintech for small business banking," or ChatGPT assembles "Toronto AI companies to watch," the systems can reach almost every company in this sample. What they find on arrival is unstructured: no FAQ schema to quote, no pricing to compare, no entity markup to anchor. For most Toronto startups, the site is reachable but the content is not built to be parsed.

Named winners: Tofu and Relay tie at 17 of 20, the highest AEO scores in the sample. Tofu is one of only two companies with FAQPage schema, and it pairs that with a valid llms.txt, Organization schema, and a hero that passes the answer-first test outright. Relay is the only company in the sample with machine-readable pricing schema, alongside full crawler access and a valid llms.txt. Clearco follows at 15 of 20.

None of this is hypothetical. The visibility gap is live: the deliberate-work criteria are where AI answers get their substance, and they sit at 29% while the access layer sits at 95%. The startups that close that gap first will inherit the citations.

Finding 2: Positioning is the one thing Toronto startups actually do well

Twenty-four of 39 companies with verifiable positioning scores (62%) scored 7.5 or higher on positioning clarity. The mean positioning score is 7.5 of 10 and the median is 8 of 10. Only one company scored below 2.5: a Toronto cybersecurity-compliance startup (a 2, due in part to domain confusion with an unrelated open-source project of the same name).

Toronto founders know how to explain what they do. This is likely a selection effect, since companies that raised money between 2024 and 2026 had to pitch clearly to get the cheque. The catch is that positioning clarity on its own does not drive discovery. A startup can ace the 10-second test and still be invisible if that positioning lives only on a homepage AI systems cannot reach and search engines do not surface.

Named winners: Eleven companies scored 9 of 10 on positioning: Beacon Software, Canada Rocket Company, Wealthsimple, Waabi, League, Una Software, Spellbook, Moonvalley, Tailscale, Tofu, and Mosaic Manufacturing. The common pattern is a specific audience plus a specific outcome, in language a competitor cannot lift. Tailscale's "Tailscale makes creating software-defined networks easy" is a line only Tailscale can own. A generic "secure networking for modern teams" could belong to anyone.

Finding 3: The content gap is severe, half the sample has no recent publishing

Nineteen of 43 companies with verifiable content scores (44%) scored 2.5 or below on content and founder-led signal, meaning no meaningful publishing in the 90 days before the audit. Eleven companies scored 0 of 10. The median content score is 5 of 10 and the mean is 4.8 of 10.

Eighteen companies scored 7.5 or higher on the content dimension: Cohere (7.5), Waabi (7.5), Float (7.5), Kepler (7.5), Ritual (7.5), Clearco (10), League (10), Ada (7.5), StackAdapt (10), Borderless AI (7.5), Tailscale (10), Tofu (10), Ecomtent (7.5), Ownright (10), Relay (7.5), PocketHealth (10), Messagepoint (7.5), Una Software (10).

Content is the compounding asset that drives organic discovery, backlinks, and inclusion in AI training data. Half of Toronto's funded startups are running on a homepage and maybe a press release. The companies scoring 10 of 10 (Clearco, League, StackAdapt, Tailscale, Tofu, Ownright, PocketHealth, Una Software) share a pattern: four or more blog posts in 90 days, visible founder LinkedIn activity, a newsletter with a real lead magnet, and press mentions linkable from the site.

Named winner: Tailscale publishes technical content at a cadence that makes it a primary source for AI systems answering networking questions. The blog does the work of infrastructure, not marketing. When an engineer asks Claude "how to set up a mesh VPN for a distributed team," Tailscale's posts are in the training data and the live retrieval results. The companies with zero blog posts in 90 days are not in that conversation.

Finding 4: Trust signals cluster at the extremes, companies have them all or almost none

Trust scores skew positive. Twenty-seven companies score 8 to 10. Nine companies score 4 or below. The mean trust score is 7.1 of 10.

The companies scoring 10: Canada Rocket Company, Smile Digital Health, Wealthsimple, Cohere, Waabi, Float, League, StackAdapt, Messagepoint, PocketHealth, Spellbook, Tailscale, Tofu. The pattern: real photos of real people, named case studies with logos, verifiable press mentions, live third-party reviews (G2, Capterra, Google), and founders identifiable as Person entities with LinkedIn profiles linked from the site.

The companies scoring 2 to 4 typically have one or two signals, usually just team photos, and nothing else. That is a strategic choice more than a resource constraint. Adding a G2 profile or linking a founder's LinkedIn costs nothing.

Finding 5: Technical health cannot be verified for the full sample

All 50 companies returned "unverifiable" for the technical health dimension because of PageSpeed Insights API rate-limiting (HTTP 429 errors) during the audit window.

This is a limitation of our infrastructure, not a startup failure. The audit tooling could not reliably measure Core Web Vitals, mobile-friendliness, HTTPS status, and sitemap submission across the full sample. For the 2027 iteration we will invest in more robust technical auditing (Lighthouse CI, dedicated PSI API quota, or similar) before collection begins. The technical health dimension is excluded from the composite this iteration.

Finding 6: Paid media presence is a black box, and the startups running ads are running everything else

Only one company showed verifiable paid media signals above 4 of 10, and that company also scored in the top quartile on content and trust. Thirty-five companies were marked "unverifiable" for paid media. Fifteen returned partial scores (0 to 4 points).

The one company with verifiable paid signals above 4 of 10 is StackAdapt (8), which scored 83 composite and 10 of 10 on content. Three more showed partial paid signals at 2 points (Ready Education, Messagepoint, Walnut Insurance), but the dataset marks most paid scores as unverifiable because transparency tools cover so little.

Paid media is genuinely hard to verify. The Meta Ad Library and Google Ads transparency tools do not capture everything, and UTM patterns are not always visible. Still, the verifiable data points one way: the companies putting money into paid acquisition are the same ones putting work into organic content and trust infrastructure. The real divide runs between full-funnel companies and underinvested ones, not between paid and organic.

Named winner: StackAdapt is the only company in the sample with verifiable paid presence above 4 of 10. It is also the number-one-rated DSP on G2. Correlation is not causation, but the pattern is instructive.

Finding 7: One in ten companies cannot be reliably fetched, and the failing set shifts week to week

Five of 50 target companies (10%) returned no scoreable data on any dimension during the May window, because the site kept failing to load. They span sports tech, event tech, e-commerce resale, consumer lending, and edtech, and they are identified in the dataset but not named here.

We then re-fetched all 50 sites in June for the AEO re-score, and the failing set moved. One of the original five, the e-commerce resale company, became reachable and earned an AEO score. Four sites that had responded in May failed in June. Four companies failed in both windows.

That movement is what tells the story. A stable outage is an infrastructure problem you fix once. A failing set that shifts between fetch windows points to intermittent bot-blocking: CDN challenge pages, rate limiting, security rules that fire unpredictably. A startup whose site is reachable this week and walled off next week is invisible to AI crawlers on exactly the days they happen to visit, and no robots.txt setting will show it.

A startup that cannot reliably serve its homepage to an auditor's fetch request cannot reliably serve it to Googlebot, GPTBot, or a customer on a slow mobile connection. Either way, one in ten of Toronto's recently funded startups has a web presence that is not consistently reachable.

Finding 8: The top companies share a specific pattern, and it is reproducible

The six highest-scoring companies (League 89, Clearco 84, StackAdapt and Tailscale tied at 83, Cohere 81, Relay 80) all score 8 or higher on positioning, 7.5 or higher on content, and 8 or higher on trust. The AEO scores below are the 20-point test normalised to 10.

• League: positioning 9, AEO 6.6, content 10, trust 10
• Clearco: positioning 8, AEO 7.5, content 10, trust 8
• StackAdapt: positioning 8, AEO 5.6, content 10, trust 10
• Tailscale: positioning 9, AEO 4.4, content 10, trust 10
• Cohere: positioning 8, AEO 6.9, content 7.5, trust 10
• Relay: positioning 8, AEO 8.5, content 7.5, trust 8

The top performers are not doing anything exotic. They have clear positioning, they publish consistently, and they build trust signals in a systematic way. What separates them is execution discipline, not strategy.

Even at the top, AEO is the weakest dimension. Tailscale, a technical company whose audience lives inside AI coding tools, captures less than half the available AEO points, with no Organization schema, no FAQ schema, and no llms.txt. Relay is the one exception, the only top-six company above 8 on AEO and the only company in the sample with machine-readable pricing. A startup that matched the top six on positioning, content, and trust and also swept the 20-point AEO test would hold a structural advantage that does not currently exist in the Toronto ecosystem.

Named winner: League's 89 of 100 is the highest score in the audit. The pattern: positioning a competitor cannot steal, full AI crawler access, a valid llms.txt, Organization schema, and trust signals across every available channel (G2, case studies, named customers, visible founders). They are the benchmark.

Finding 9: Deep tech content splits cleanly into publishers and non-publishers

Companies in biotech, aerospace, and quantum-adjacent deep tech show a clean split within the group on content. Waabi and Kepler Communications both score 7.5 of 10 because they publish. The pure-science subset (a rocketry startup and two biotechs) averages 4.2 of 10, below the 4.8 of 10 sample mean. The sharpest case in the group pairs a perfect 10 of 10 on trust with 2.5 of 10 on content: real credentials, no distribution.

Deep tech companies have real credentials: press coverage, named team members, scientific advisory boards, patents. The non-publishers underinvest in the content that would make those credentials discoverable. A biotech with strong publications and no blog is invisible to the AI systems answering "best Canadian biotech for oncology immunotherapy." The trust is earned. The distribution is not built. Waabi and Kepler are the counter-examples: both publish insights content that translates technical work into plain language without dumbing it down.

Named winner: Waabi scores 7.5 on content despite being a physical AI company building autonomous trucking systems. Its "Insights" section turns research into accessible writing without losing the technical rigour, which is the model other deep tech companies should copy.

Finding 10: The median Toronto startup scores 59 of 100, passing but not competitive

The median composite across the 44 companies with verifiable composite scores is 59 of 100. The mean is 56.5 of 100. The distribution concentrates in the 51 to 75 band, with a high-performing cluster of eight companies at 76 to 100. The range runs from 0 (a Toronto proptech) to 89 (League).

Distribution by band:

• 0-25: 4 companies
• 26-50: 11 companies
• 51-75: 21 companies
• 76-100: 8 companies

A 59 of 100 is a C-plus. It means the median Toronto startup has clear positioning, decent trust signals, and AEO readiness that arrived free with a modern web framework, and not much beyond that. Inconsistent content. Unverifiable technical health. Unverifiable paid presence.

The gap to the 76 to 100 band is not incremental. Those companies run a different playbook, and the distance between 59 and League's 89 is the distance between having a website and having a discovery engine.

For founders reading this: score above 75 and you sit in the top fifth of the Toronto funded startup ecosystem. Score below 50 and you are in the bottom third, while the companies above you compound and you do not.

Ranked tables and distribution

Top 10 companies by composite score

Note: AEO is shown as the raw 20-point test score; it enters the composite normalised to /10. Technical health and paid media dimensions are omitted from the table because of high unverifiable rates (100% and 70% respectively); where paid media was verifiable it still counts toward the composite, which is why Tofu (paid 0/10) and StackAdapt (paid 8/10) rank where they do. On AEO alone, Tofu and Relay lead the entire sample at 17/20.

Median scores by dimension

The headline: Toronto startups can explain what they do, and AI systems can reach them. What those systems find on arrival is unstructured, and the substance that earns citations is missing from seven sites in ten.

What this means for Toronto startup founders

If you raised funding in the last two years and your composite is below 50, you are in the bottom third of the ecosystem, and the companies above you compound their discovery advantage every quarter. Here is what to fix first.

If your AEO score is low, run the 20-Point AEO Readiness Test on yourself. Eight checks, 2.5 points each, every one verifiable in under five minutes:

1. Organization schema. JSON-LD on your homepage identifying who you are. Fifteen minutes.
2. FAQPage schema. On your top three pages, with real questions your customers ask. An afternoon.
3. Answer-first hero. The first sentence on your homepage states what you do and for whom. No slogan survives this test.
4. AI crawler access. robots.txt allows GPTBot, CCBot, Google-Extended, PerplexityBot, and ClaudeBot. Then check your CDN settings, because Cloudflare can block AI bots even when robots.txt says yes.
5. llms.txt. A markdown index of your key pages at yourdomain.com/llms.txt. An emerging standard rather than a ranking factor, but AI coding and research agents fetch it directly, and it costs an hour.
6. Render without JavaScript. Load your homepage with JavaScript disabled. If it is blank, you are blank to most AI crawlers, because they do not run JavaScript. This is the single most damaging failure on the list.
7. Pricing transparency with schema. Product or Service schema with price fields, or at least a pricing page that declares your model. AI systems answering "how much does X cost" skip the companies that hide it.
8. Entity alignment. Your LinkedIn company page, Crunchbase profile, and website all live, all pointing at the same domain, all describing the same company. AI systems cross-reference entities, and inconsistency reads as uncertainty.

Score yourself honestly. The median funded Toronto startup fails six or more of these. Most are hours of work, not quarters.

If your content score is below 5: commit to one blog post a week for the next 90 days. The posts do not need to be long. They need to be specific, answer-first, and tied to a real customer question. If you do not know what to write about, open your support inbox: a support ticket, a sales objection, or a feature launch each makes a post. The companies scoring 10 of 10 on content win on discipline, not creativity.

If your trust score is below 6: add real photos of your team to the About page, add named case studies with customer logos to the homepage, link your founders' LinkedIn profiles from the About page, and create a G2 profile with reviews from your three happiest customers. Two days of work.

If your positioning score is below 7: run the 10-second test. Show your homepage to someone who has never heard of your company, count to ten, and ask them to say who you serve and what makes you different. If they cannot, your positioning is not clear. Rewrite the hero until a first-time visitor passes. Positioning is the foundation everything else compounds on.

The startups scoring 75 and up are not doing anything you cannot do. They do the boring work consistently, and that consistency is the whole gap.

What this means for Toronto VCs and ecosystem operators

The startups in your portfolio with composite scores below 50 are not GTM-ready. They may have product-market fit, they may have revenue, but they do not have a discovery engine. When a potential customer searches their category, they do not appear. When an AI system generates a list of companies to evaluate, they are not on it.

The good news is that this is fixable. The companies scoring 75 and up are running a reproducible playbook, not sitting on an advantage others cannot copy. If you are a VC, consider making AEO readiness, content cadence, and trust-signal accumulation part of your post-close onboarding checklist. If you are an ecosystem operator, consider offering AEO and content strategy as a standard service to portfolio companies.

The companies that fix this in 2026 will compound their discovery advantage through 2027 and 2028. The ones that do not will fall further behind every quarter.

Limitations

Technical health dimension: the audit infrastructure could not reliably measure Core Web Vitals, mobile-friendliness, HTTPS status, and sitemap submission across the full sample because of PageSpeed Insights API rate-limiting (HTTP 429 errors) during the audit window. For the 2027 iteration we will invest in automated technical auditing (Lighthouse CI, dedicated PSI API quota) before collection begins. The technical health dimension is excluded from the composite this iteration.

Paid media dimension: paid media is genuinely hard to verify. The Meta Ad Library and Google Ads transparency tools do not capture all activity, and UTM patterns are not always visible. Treat the paid media findings as directional rather than definitive. Only 15 of 50 companies (30%) returned verifiable paid media scores, and 35 were marked unverifiable.

Sample bias: the sample is filtered to companies that raised funding between January 2024 and April 2026, which introduces selection bias, since funded companies are likely better at positioning and trust signals than companies that did not raise. The findings do not generalise to the broader Toronto startup ecosystem, only to the funded subset.

Single-reviewer scoring: each company was scored by a single human reviewer to keep scoring consistent, which introduces reviewer bias. For the 2027 iteration we will implement dual-reviewer scoring with a tie-breaker protocol for disagreements.

Fetch failures: four of 50 target companies could not be audited on any dimension in either fetch window, and several more only partially, because of site fetch failures. The failing set shifted between the May window and the June AEO re-fetch, so some failures are intermittent bot protection rather than stable outages. We cannot separate bot protection from a genuine infrastructure problem without manual outreach, which was outside the scope of this audit. For that reason, unscored companies are identified in the dataset but not named in this report.

CDN-level bot blocking: a site can allow every AI crawler in robots.txt and still be unreachable to them, because CDN defaults (Cloudflare's AI-bot blocking, challenge pages, rate limiting) operate below the robots.txt layer. Our robots.txt criterion measures stated policy, and our fetch results measure actual reachability. Where the two disagree, the fetch result is the one AI systems experience.

Frequently asked questions

How were the 50 companies selected?

We compiled a list of Toronto and GTA-headquartered startups that publicly announced a funding round between January 2024 and April 2026, using BetaKit funding announcements, the Crunchbase Toronto/GTA filter, the Communitech member directory, and the MaRS portfolio. We excluded stealth-mode companies with no website, companies acquired or wound down before the audit date, and companies headquartered outside the GTA even if Toronto-incubated.

Why did some companies fail to load?

Site fetch failures can come from aggressive bot protection (Cloudflare challenge pages, rate limiting), temporary downtime during the audit window, or genuine infrastructure problems. We attempted multiple fetches per company across two windows (May and June 2026), and the set of failing sites shifted between them, which is evidence that much of the blocking is intermittent rather than structural. Where no fetch succeeded, we marked the affected dimensions unverifiable rather than guessing.

Can I see the underlying dataset?

Yes. Email dave@stratezik.com for access to the full dataset (CSV format) with company names, scores, and notes. We will also provide sector-specific cuts on request (for example, fintech only or B2B SaaS only).

Will you audit the same companies in 2027?

Yes. The same 50 companies will be re-audited in Q2 2027 using the same rubric. Year-over-year comparison will become the headline finding in the 2027 report.

My company is in the dataset and I disagree with the score. What do I do?

Email dave@stratezik.com with the specific dimension you are disputing and supporting evidence (screenshots, links, third-party verification). If the evidence is conclusive, we will update the score and republish the dataset with a changelog note. If it is inconclusive, we will note the dispute in the dataset and leave the score unchanged.

Can I use this data in a presentation or article?

Yes. Attribute it to "Stratezik Toronto Startup Website Audit 2026" and link to this report. If you are a journalist or researcher, email dave@stratezik.com for embargoed access to the 2027 report.

Methodology appendix and dataset access

The full dataset (CSV format) includes company name, domain, six dimension scores, composite score, and notes for each of the 50 target companies. The dataset holds 50 company rows across 6 dimension columns, or 300 individual data points.

Request access: email dave@stratezik.com with the subject line "Toronto Startup Audit 2026 Dataset Request." Include your name, organisation, and intended use. We will respond within 48 hours with a download link.

Sector-specific cuts: if you need a filtered view (fintech only, B2B SaaS only, deep tech only), say so in your request and we will provide a custom CSV.

Changelog: if any scores are updated after publication because of founder disputes or new evidence, we will publish a changelog at the bottom of this page with the date, company name, dimension, old score, new score, and reason.

Sources

1. BetaKit funding announcements (January 2024 to April 2026): betakit.com/tag/funding
2. Crunchbase Toronto/GTA filter: crunchbase.com/hub/toronto-startups
3. Communitech member directory: communitech.ca/members
4. MaRS portfolio list: marsdd.com/portfolio
5. Stratezik internal audit dataset (May and June 2026): available on request at dave@stratezik.com
6. SE Ranking, llms.txt adoption study across 300,000 domains (2026): seranking.com/blog/llms-txt-study
7. Google's position on llms.txt, from John Mueller (Search Off the Record and public statements) and Gary Illyes (July 2025): Google does not use llms.txt. Summary: presenc.ai/research/state-of-llms-txt-2026

---

About Stratezik: Stratezik is a Toronto-based marketing agency that runs on its own AI agent system. We specialise in AEO-first content strategy, founder-led brand building, and full-funnel paid media for startups and scale-ups. Stratezik's paid media lead has managed $10M+ in annual ad spend and is a Google Search Honours Award recipient, and its brand lead brings 15+ years of brand marketing experience including general management.

Contact: dave@stratezik.com