How the scores work
Everything on this site comes from measured data collected on 23 September 2026, for Google US in English. Here's what we measured, how it turns into a score and where the limits are.
What we measured
- Keywords. Search volume, keyword difficulty, CPC, intent, AI search volume and the year-on-year trend for 3,009 keywords: each brand, "[brand] review", "alternatives", "pricing" and "affiliate program", plus a head term per category (DataForSEO Labs and AI Keyword Data).
- Live results pages. The Google top 10 for 1,776 searches, with AI Overview text and its cited sources, People also ask and related searches. Every result is classed as vendor, rival vendor, review platform (G2, Capterra, Trustpilot and the like), forum or video, directory or independent.
- AI mentions. How often each brand appears in Google AI Overviews and ChatGPT answers, and which sources those answers cite (DataForSEO LLM Mentions).
- AI answers. 479 full answers from ChatGPT and Gemini with web search on: "best X" and "best X affiliate programs" for every category, five general affiliate questions and "Is X worth it?" for 153 programs.
- Commission terms. The 119 board programs have hand-checked terms. Every off-board program that claimed 30% or more recurring was verified against its own affiliate page or portal on 23 September, and on 26 September we re-read the vendor pages of 49 more high-traffic programs whose terms had been collected automatically. Everything else is marked "unverified", and we don't repeat those figures in titles, answers or summaries.
- AI demand estimates. The "Est. AI demand" figures come from DataForSEO's AI Keyword Data. It is a modeled estimate based partly on Google's People also ask data, not a count of prompts anyone observed.
Winnability (0 to 100)
- Demand, up to 30 points: log of review searches + 0.6 x alternatives + 0.4 x pricing.
- Difficulty, up to 20: the inverse of volume-weighted keyword difficulty.
- Openness, up to 25: independent sites in those top 10s. Five or more gets full marks.
- AI opportunity, up to 15: half AI mention volume, half the share of AI citations that go to independent sites.
- AI Overview penalty, down to -10: the share of those searches that show an AI Overview.
- Demand gate: the whole score is halved under 50 weighted searches and multiplied by 0.8 under 150, so an empty niche can't score well just because it's uncontested.
Content opportunity bands: A, strong content bet at 60+, B, worth a post at 45 to 60, C, roundup only at 30 to 45, D, skip as a page under 30 and no search demand under 10 weighted searches. These rate the chance to win search traffic. They are not a recommendation to join the affiliate program: check the separate terms status for that.
Index vs GoHighLevel (board only)
35% economics (the log of the 24-month referral value, discounted 25% when terms or price aren't fully confirmed), 50% winnability and 15% program friction, all multiplied by the same demand gate. It's scaled so GoHighLevel, the benchmark, is 100.
24-month referral value
Monthly commission x the sum of (1 - churn) to the power t over 24 months, or fewer if the program caps payouts. Churn is 3% a month for sticky tools, 5.5% for average ones and 8.5% for tools people drop quickly. These rates are our modeling assumptions based on the type of tool, not retention measured for each program, so every 24-month value on this site is an estimate.
Known limits
- US English only, so programs with mostly international demand look weaker than they are.
- "Independent" means a publisher: not the vendor, a review platform, a forum, a directory or another software company. Since the 2 November 2026 data update we review every domain the automatic rules would call independent. Of the 3,064 such domains, 2,127 turned out to be software or service companies' own sites (for example, rival hosts ranking for "Kinsta alternatives") and now count as rival vendors, and 174 are directories or marketplaces. That stricter definition lowered most winnability scores. Some of these calls are judgment calls, so treat a single classification as a signal rather than a fact.
- AI answers vary from run to run. Treat a single mention as a signal, not a fact.
- Ambiguous brand names were searched with a qualifier (Kit as "convertkit", Close as "close crm" and so on).
- Terms change. Every program page shows the date we checked and links to the source.
Total data cost for this edition: $14.07 in DataForSEO API calls. The full dataset is on the data page.