9,550 AI-built startup entries from 3 public builder contests (as of 2026-09-20). Entries submitted to more than one contest are counted once. Themes are keyword labels on pitch text; one entry can carry several. Traction and other signals are what builders wrote, not verified facts. "Under-represented in this dataset" describes these entries only — it is not an untapped market or market size.
The dataset holds 9,550 unique AI-built startup entries from 3 public builder contests (as of 2026-09-20).
How: unique products after cross-source deduplication · denominator: all loaded triaged rows
Limits: snapshot date is the latest submission 'Created' value, a proxy rather than an extraction date
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
751 of 10,301 submitted rows were duplicates of another entry; 693 products appear in more than one contest.
How: rows_loaded - unique_products · denominator: rows_loaded
Limits: duplicate detection is rule-based (see stats.dedup_rule); some resubmissions under new wording may be missed
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
34.7% of entries (3,312 of 9,550) have their own custom domain.
How: count / unique products · denominator: unique products (combined)
Limits: has a custom domain not on a shared hosting platform; NOT evidence of commercial maturity
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
India-contest entries: 30.8% have their own custom domain; global-contest entries: 38.9%.
How: count / unique products per regional subset · denominator: unique products per regional subset
Limits: has a custom domain not on a shared hosting platform; NOT evidence of commercial maturity; region is the contest's grouping, not the product's market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
2.0% of entries (191 of 9,550) mention any traction (users, customers, revenue) in their pitch text.
How: count / unique products · denominator: unique products (combined)
Limits: description/title text matches a traction pattern in themes.yaml; a TEXT SIGNAL, not verified revenue or demand
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
India-contest entries: 2.2% mention any traction (users, customers, revenue) in their pitch text; global-contest entries: 2.1%.
How: count / unique products per regional subset · denominator: unique products per regional subset
Limits: description/title text matches a traction pattern in themes.yaml; a TEXT SIGNAL, not verified revenue or demand; region is the contest's grouping, not the product's market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
0.3% of entries (24 of 9,550) say they were built for or used in the builder's own business.
How: count / unique products · denominator: unique products (combined)
Limits: text says it was built for/used in the builder's own business; self-reported
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
India-contest entries: 0.4% say they were built for or used in the builder's own business; global-contest entries: 0.1%.
How: count / unique products per regional subset · denominator: unique products per regional subset
Limits: text says it was built for/used in the builder's own business; self-reported; region is the contest's grouping, not the product's market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
5.3% of entries (502 of 9,550) are all-in-one business suites or billing/accounting tools.
How: count / unique products · denominator: unique products (combined)
Limits: carries erp_suite or billing_accounting theme label
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
India-contest entries: 7.4% are all-in-one business suites or billing/accounting tools; global-contest entries: 4.1%.
How: count / unique products per regional subset · denominator: unique products per regional subset
Limits: carries erp_suite or billing_accounting theme label; region is the contest's grouping, not the product's market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'Education & edtech' is crowded in this dataset: 761 entries (8.0%), rank 1 of 49 themes.
How: unique products carrying theme label / unique products · denominator: 9,550 unique products (combined)
Limits: keyword classification; multi-label; crowded in submitted entries, not a statement about the whole market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'Personal finance, investing & fintech' is crowded in this dataset: 718 entries (7.5%), rank 2 of 49 themes.
How: unique products carrying theme label / unique products · denominator: 9,550 unique products (combined)
Limits: keyword classification; multi-label; crowded in submitted entries, not a statement about the whole market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'Data, analytics & BI' is crowded in this dataset: 611 entries (6.4%), rank 3 of 49 themes.
How: unique products carrying theme label / unique products · denominator: 9,550 unique products (combined)
Limits: keyword classification; multi-label; crowded in submitted entries, not a statement about the whole market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'AI agents & workflow automation' is crowded in this dataset: 418 entries (4.4%), rank 4 of 49 themes.
How: unique products carrying theme label / unique products · denominator: 9,550 unique products (combined)
Limits: keyword classification; multi-label; crowded in submitted entries, not a statement about the whole market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'Billing, invoicing & accounting' is crowded in this dataset: 354 entries (3.7%), rank 5 of 49 themes.
How: unique products carrying theme label / unique products · denominator: 9,550 unique products (combined)
Limits: keyword classification; multi-label; crowded in submitted entries, not a statement about the whole market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'Logistics, fleet & delivery' is crowded in this dataset: 347 entries (3.6%), rank 6 of 49 themes.
How: unique products carrying theme label / unique products · denominator: 9,550 unique products (combined)
Limits: keyword classification; multi-label; crowded in submitted entries, not a statement about the whole market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'CRM, sales & lead management' is crowded in this dataset: 345 entries (3.6%), rank 7 of 49 themes.
How: unique products carrying theme label / unique products · denominator: 9,550 unique products (combined)
Limits: keyword classification; multi-label; crowded in submitted entries, not a statement about the whole market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'Restaurants, food & hospitality' is crowded in this dataset: 332 entries (3.5%), rank 8 of 49 themes.
How: unique products carrying theme label / unique products · denominator: 9,550 unique products (combined)
Limits: keyword classification; multi-label; crowded in submitted entries, not a statement about the whole market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'All-in-one business suite / ERP' is relatively under-represented among global-contest entries (1.2%, 73 entries) versus India-contest entries (2.9%, 122).
How: share_low <= 0.5 * share_high and count_high >= WHITESPACE_MIN_CLUSTER · denominator: unique products per regional subset
Limits: relative under-representation in this dataset only; region is the contest's grouping, not the product's market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'Manufacturing & factories' is relatively under-represented among global-contest entries (1.1%, 67 entries) versus India-contest entries (2.6%, 113).
How: share_low <= 0.5 * share_high and count_high >= WHITESPACE_MIN_CLUSTER · denominator: unique products per regional subset
Limits: relative under-representation in this dataset only; region is the contest's grouping, not the product's market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
'Agriculture & farming' is relatively under-represented among global-contest entries (0.9%, 50 entries) versus India-contest entries (2.0%, 86).
How: share_low <= 0.5 * share_high and count_high >= WHITESPACE_MIN_CLUSTER · denominator: unique products per regional subset
Limits: relative under-representation in this dataset only; region is the contest's grouping, not the product's market
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
Among 66 shortlisted finalists in one contest, 23 had a dead or unreachable product page at evaluation time.
How: count of DEAD class / finalists · denominator: finalists in source
Limits: historical observation at the original evaluation date; a dead page is an access signal, not proof of business failure; site status free text classified by documented keyword rules
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1
Among 66 shortlisted finalists in one contest, 14 had a dead or unreachable product page at evaluation time.
How: count of DEAD class / finalists · denominator: finalists in source
Limits: historical observation at the original evaluation date; a dead page is an access signal, not proof of business failure; site status free text classified by documented keyword rules
Snapshot 2026-09-20 · corpus corpus-7a0761e3fa14 · themes-v1