THE DATA under 'match the TLD to the buyer' (3-sweep research 2026-08-27; confidence graded, facts-not-opinion). HEADLINE CORRECTION: the reason .com matters for a money/mainstream buyer is NOT trust-at-checkout -- that is FOLKLORE. No controlled study in 15+ yrs of CRO isolates TLD as a checkout/conversion variable; Baymard (44k hrs) + Nielsen Norman (350+ sites) do NOT list TLD among trust signals (design, security badges, reviews, price dominate). Do NOT justify .com with 'users won't pay on a non-.com'. The 3 REAL, MEASURED costs of a non-.com: (1) RECALL + TYPE-IN LEAKAGE [Med] -- .com correctly recalled 44% vs 25-33% for alts, and when people misremember a TLD they type .com 3.8x more than any other guess (GrowthBadger n=1,500) -> word-of-mouth + direct-navigation traffic LEAKS to whoever owns your .com. Matters for recall/WOM-driven acquisition -- but that channel is SHRINKING (zero-click 58.5%, AI-Overviews cut organic CTR ~61%) so the cost is real yet bounded. (2) ABUSE-DRIVEN DISTRUST + EMAIL DELIVERABILITY [High] -- new gTLDs run ~0.6% abuse vs 0.019% for legacy .com/.net/.org = 32x (DNSRF 2025); phishing-domains/10k: .com 18.5, .top 99.5, .xyz 37.8, .icu 555, .cfd 487 (Interisle 2025-26); junk TLDs (.zip/.mov/.xyz/.top/.sbs/.cfd/.icu/.cc) are BLOCKED by mail servers + enterprise security as standing policy (Cisco/Microsoft/SANS) -> a legit brand on a junk TLD inherits the distrust + a DELIVERABILITY tax (matters for his drip/email ventures). MECHANISM = cheap+frictionless registration, NOT the letters (Brand TLDs = 0% abuse) -> which is exactly why .ai/.io are OFF every abuse list (real registry, ~$70+/yr friction) and must NOT be lumped with junk. (3) SHARED-NAMESPACE + REGISTRY-EXISTENCE RISK [High] -- a small/abused ccTLD blacklists your domain via OTHERS' behavior (Notion's .so got malware-blacklisted -> bought notion.com 2026); .io is on a 5-10yr IANA RETIREMENT CLOCK (Chagos->Mauritius ISO-code loss); .ai = a ccTLD you rent from Anguilla; .ly = Sharia-takedown precedent. --- THE DECISION MATRIX (STAKES axis > AUDIENCE axis, both evidenced): STAKES/CATEGORY = the STRONGER predictor [High] -- the more transactional/regulated/mainstream-consumer the buyer, the harder .com dominance (fintech 88% .com, all-funded 67.9%, YC/technical only ~46-53%). AUDIENCE = the WEAKER predictor [Med] -- technical/dev/AI-native buyers demonstrably tolerate alts (YC/Techstars 28% .ai) and ~25% of mainstream consumers distrust .ai/.io (worse among older/women, better among 18-34/men; single vendor survey).
CRITICAL PATTERN [High]: STAKES OVERRIDES AUDIENCE once money moves -- Cursor(.sh), Perplexity(.ai), Notion(.so), OpenAI all SCALED on an alt-TLD then BOUGHT the .com at Series-B/scale; NOBODY goes .com->alt. So LAUNCHING on the best-name clean alt-TLD is fine for a technical/AI-native/low-stakes product; the .com upgrade is a RITE OF PASSAGE when it turns transactional/enterprise/mainstream (cheap now for most, ~$1-4k; up to 6 figures for a hot exact-match). BONUS: .dev/.app are HSTS-preloaded (Google-run, HTTPS-forced) = a security-POSITIVE for a dev tool. NET RULE (his 'money buyer vs just-for-fun' intuition, corrected + sharpened): money-on-the-line / mainstream / regulated buyer -> .com or <name>qualifier.com for RECALL + ANTI-LEAKAGE + CATEGORY-FIT (NOT mythical checkout-trust); low-stakes / technical / AI-native -> best NAME wins on a CLEAN whitelisted alt (.ai/.io/.dev) -- but a JUNK high-abuse gTLD is NEVER a 'just for fun' pass, it is a deliverability + security cost.