TLDs & the aftermarket

SUPERSEDES the old 'ABSOLUTE .com-only FOREVER' rule -- the owner RELAXED it 2026-08-27 to fit what works now (.com fell to ~46% of YC startups, Placek calls .com just an area code, and serious AI cos SHIP AND STAY on .ai). WE ARE NOT BEING LAX -- we are LEAD NAMERS with a considered, ORDERED TLD hierarchy. The NAME-quality bar (GATE 0) is UNCHANGED; this governs ONLY the TLD once a great name exists, and it is BUYER-MATCHED (see com_by_buyer_not_fashion). RULE: take the HIGHEST-RANKED whitelisted TLD the great name is available on, for THAT buyer. --- WHITELIST (strongest -> situational): (1) <name>.com -- the DEFAULT for any brand; still the plurality + highest type-in trust + the only extension a mainstream/SMB/non-technical buyer fully trusts. (2) <name><qualifier>.com (tryprofound.com, novalabs.com, onhandos.com -- brand word leads, qualifier hq/labs/get/try/os/co follows) -- the PREFERRED move when the exact-match .com is taken; keeps the .com trust anchor + full ownership; RANKS ABOVE any alt-TLD. The buy-path is now CHEAP (~$1-4k, see AFTERMARKET_PLAYBOOK) so an aftermarket <name>.com often beats a bare alt-TLD. (3) <name>.ai -- FIRST-CLASS + category-POSITIVE for an AI-NATIVE brand with an AI-literate audience (Claude.ai/Perplexity.ai/Mistral.ai/xAI/Character.ai stay on .ai at multi-B valuations; .ai passed 1M regs Jan-2026). Caveat: ccTLD (Anguilla) = you rent a country politics + price rising ($70->$80); use for AI PRODUCTS, NOT a mainstream/SMB brand or the holding. (4) <name>.io -- TOLERATED ONLY for a dev/technical-audience tool, never a flagship/holding. FLAG: Chagos/Mauritius sovereignty tail-risk (treaty shelved early-2026, currently stable; a ccTLD retirement would run 5-10yr if ever triggered). (5) <name>.co -- WEAK, last-resort only (reads couldnt-get-the-.com to most buyers); technical buyer + strong brand only. --- BLACKLIST (never the LIVE brand): hyphenated .com; suffix-MANGLE shown as the brand (suffix-mangle tell); .xyz / .org / .net / .cash / .app / novelty gTLDs as the primary brand (weak trust / spam-adjacent / wrong register for his premium positioning); any TLD that makes the phone test fail. (.net/.org OK as DEFENSIVE redirects only -- see defensive_set.) --- DECISION: match TLD to BUYER, then take the highest-ranked whitelisted one available. Taken .com for a MAINSTREAM buyer -> a <name>qualifier.com or a different NAME (NOT a bare weak TLD); for an AI-NATIVE product, .ai is a legitimate FIRST choice, not a downgrade. Do not revert to the old absolutism, and do not over-swing into any-TLD-goes -- the RANKING is the discipline.

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.

LIVE AGAIN (the old SUPERSEDED tag removed 2026-08-27 when the absolute rule was relaxed) -- this buyer-match logic is now the governing principle under RANKED_TLD_POLICY. .com = ~46% of YC startups (2025) but still the plurality + highest RECALL (44% correctly recalled vs 25-33% for alts) + the default type-in guess (3.8x) so it leaks the least word-of-mouth traffic, and the category norm for mainstream/regulated/consumer buyers (fintech 88% .com) -- NOTE 'trust-at-checkout' is FOLKLORE, do NOT cite it; see BUYER_MATCH_EVIDENCE for the graded data. So: mainstream/SMB/luxury/holding buyer -> .com (or <name>qualifier.com); AI-native product to an AI-literate buyer -> .ai is first-class; dev tool -> .io tolerated (Chagos-flagged). The TLD is chosen from the BUYER, not fashion and not an absolute.

Hyphenated .com (cheapness signal + leaks type-ins to the squatter); suffix-mangle domains shown as the LIVE brand (suffix-mangle tell that you lost the clean name); settling for a WEAK/blacklisted TLD (.xyz/.co/etc.) just because the .com is taken -- for a mainstream buyer pick a different NAME or the <name>qualifier.com. A whitelisted alt-TLD matched to the buyer (.ai for an AI product) is NOT settling.

Per RANKED_TLD_POLICY (2026-08-27). Ordered whitelist: .com > <name>qualifier.com > .ai (AI-native only) > .io (dev tool only, Chagos-flagged) > .co (weak, last resort). Blacklist: hyphenated .com, suffix-mangle-as-brand, .xyz/.org/.net/.cash/.app/novelty as the live brand. .net/.org = defensive redirect only.

The LIVE BRAND is the highest-ranked whitelisted TLD per RANKED_TLD_POLICY (usually .com; .ai for an AI-native product) -- everything else here is DEFENSIVE PARKING only. Fine to register a few typo variants (301 -> the live domain). Parking a .net/.org purely to redirect for anti-phishing is allowed ONLY once there is a brand worth protecting AND it never shows as the live brand -- do NOT reflexively grab them, and NEVER let a parked .net/.org drift into being the brand. Default: the live domain + a typo or two. Math: $10/variant vs UDRP $1,500+ vs litigation $10k+.

CORRECTED 2026-08-27 [High] — the old '$23,264 average / $3k-15k brandable sweet spot' was WRONG (a mean-of-a-power-law + vendor inflation; it made BUYING look ~5x costlier than reality and wrongly closed the buy path). GROUND TRUTH = the largest real dataset, Bishopi's 480,667 .com sales Jan-2024→May-2026 ($491M total): 99.4% of ALL .com sales close UNDER $10,000. Distribution: <$100 = 31% (median $57) · $100-999 = 56% (median $255) · $1k-9,999 = 12% (median $1,975) · $10k-99k = 0.6% (median $15,508) · $100k+ = only 0.04% (median $210k). Length premium: 3-char median $1,425, 4-char $234, 5+ char flattens to ~$180-195. SO a decent one-word brandable .com realistically trades LOW-TO-MID FOUR FIGURES ($1,000-$4,000); a genuinely rare dictionary word / sub-4-char = five figures; six figures = the top 0.04%, NOT typical. Venues (2026): GoDaddy = volume/sub-$2k speed (avg ~$778) · AFTERNIC = mid-market $2k-10k (avg ~$3,879) · SEDO = premium $5k+ / broker channel (avg ~$13,916) · ATOM (ex-Squadhelp) = brandables specialist (15-35% commission). ⚠️ Dan.com is DEAD (GoDaddy folded it into Afternic end-2024; escrow commission rose 9%→25%). Afternic floors sit ~20-30% below Buy-Now → open at 60-70% of ask. Sell-through ≈1-3%/yr → time is the buyer's leverage. Many great .coms are 'held, not listed' → broker-outreach (anonymous, patient), not an instant buy.

NEVER inquire from a company email or reveal launch/funding — corporate identity = 5-10x ask. Personal email, patient, unbothered, cite comps (NameBio-style), never reveal budget. Walk-away is the whole leverage (same law as his dating doctrine). Always Escrow.com (0.89-3.25%) for direct deals; independent commission-only brokers > GoDaddy DBS ($99.99+20%) for stealth.

pendingDelete mechanics: expiry → grace ~0-45d → redemption ~45-75d → pendingDelete 5d frozen → drop (Verisign deletes ~11:00-15:00 UTC). DropCatch: free to place, $59 only on success, 60-80% catch on moderately-competitive names. TRIPLE-BACKORDER (DropCatch+NameJet+Dynadot) = ~35% higher combined catch. Multiple backorders → private auction: set a max ceiling BEFORE, don't chase.

(ecom counterweight) Paid premium is RATIONAL for a retainer/service business (name amortizes over $10k/mo clients) and was correctly REJECTED for unproven DTC brands (ecom passed on $895 adriata) — spend scales with the venture's validated stage, never before WTP.

The free .com pool for good names is ~EXHAUSTED. ~750 candidates checked 2026-07-03..05: ALL real dictionary words, ALL first-order A1 compounds (gold/top/star/king/echo/spark × word/talk/rank/say...), ALL iconic phrases (namedrop, pageone, lastword, hearsay...), ALL Latin single words (Teneo-class) = registered or parked ($20k–$39M asks).

RE-CONFIRMED 2026-07-06 (~2,250 MORE .com checks: single words, obvious compounds, clever hand-crafts, 1,000+ builder-compounds): every name that CLICKS is taken; every FREE one is mush (prowroar/keelleap/answermade). The free-real-word-.com door does NOT exist. The ONLY free+great+.com lane = the DISTANT-LEAP BLEND (GENERATION_METHODS.distant_leap_blend_PRIMARY); great real words = BUY (a good brandable is ~$1-4k, NOT the inflated '$23k avg' the file used to cite — corrected via the 480k-sale Bishopi dataset, see AFTERMARKET_PLAYBOOK).

CORRECTED 2026-07-06. The old 'second-order roots / -won seam / small aftermarket' lanes yield MEDIOCRE, invoice-failing names (three of them are literally in a graveyard) — do NOT mine them for a GREAT name. For GREAT + free + .com there is exactly ONE lane: (1) the DISTANT-LEAP BLEND — two real words from far-apart domains (the ByteDance move; GENERATION_METHODS.distant_leap_blend_PRIMARY). Otherwise BUY or CATCH: (2) two-word brandable ~$1-4k (Atom/Afternic — the old '$3-15k / ~$23k avg' was inflated, corrected 2026-08-27, see AFTERMARKET_PLAYBOOK); (3) DROP-CATCH a great expiring .com — DropCatch (60-80% catch on moderate names, ~$59-70) + NameJet in parallel (~35% higher combined), private auction if contested (set a ceiling first). NOTE: cheap ≠ automatically mediocre — many GOOD brandables sit in the low four figures; judge by GATE 0, not by price.

Open seams close in DAYS once discoverable (drumgold + bravenzo registered days before the hunt found their lanes). Verified-available = register same-day or lose it.

Any 'available' list in this file is a SNAPSHOT (2026-07-05) — re-whois before acting.