Methodology
RecoIndex measures one thing: when someone asks an AI assistant for a product recommendation, which brands does it name? We run a fixed, public set of buyer-intent questions across several models, repeat them, and report how often each brand appears — with sample sizes and the sources the models cited. Everything here is reproducible: ask the same questions yourself and you should see broadly the same pattern.
1. The questions
Each category uses the same set of buyer-intent prompt templates — the questions people actually ask before choosing a product — instantiated with that category's product noun. We do not steer the model toward any brand. The current template set:
- What is the best accounting software for a small business?
- Which accounting software should a 10-person startup use?
- Recommend a accounting software for a B2B sales team.
- What accounting software do you recommend for a bootstrapped SaaS company?
- I need a accounting software that's easy to set up. What should I pick?
- What's the most affordable accounting software that's still good?
- What is the best accounting software for a growing mid-size company?
- Which accounting software has the best free plan?
- What accounting software would you choose in 2026, and why?
- What are the top 5 accounting software options right now?
- What's a good alternative to the most popular accounting software?
- Which accounting software is best for a fully remote team?
- Recommend a accounting software with the best integrations.
- Which accounting software should an agency use to manage multiple clients?
- What is the best enterprise-grade accounting software?
- Which accounting software do you recommend for a solo founder?
- Which accounting software offers the best value for money?
- If you had to pick just one accounting software today, which would it be?
- Which accounting software is best for a company that cares about data privacy?
- Which accounting software is easiest to migrate to from a spreadsheet?
2. The models
We sample these models via OpenRouter:
- Claude Sonnet 4.6 (anthropic/claude-sonnet-4.6)
- DeepSeek V3.1 (deepseek/deepseek-chat-v3.1)
- Gemini 2.5 Flash (google/gemini-2.5-flash)
- ChatGPT (GPT-5.4) (openai/gpt-5.4)
3. Sampling
Language models are non-deterministic — ask twice, get different answers. So we never report a single answer. Each (question × model) is sampled multiple times at a fixed temperature, and we report mention frequency with the sample size shown (e.g. "named in 18 of 24 answers"). A second, low-cost model extracts the brand names and any cited URLs from each answer into structured data; that extraction is deterministic (temperature 0).
4. What we report
- Mention rate — share of sampled answers that named the brand.
- First-pick rate — share of answers that named it first.
- Cited sources — URLs the models referenced, ranked by frequency.
5. Limitations
- We query models via API, which can differ from the consumer chat apps (system prompts, browsing, personalization).
- Results are not personalized — no location, history, or account context.
- Brand-name normalization is automated; near-duplicates (e.g. "HubSpot" vs "HubSpot CRM") are merged heuristically and refined over time.
- Sample depth grows week over week; early weeks have smaller samples (shown as the n on each page).
- Models and the web they cite change constantly — this is a moving picture, not a verdict.
6. Raw data
Every category page has a "Download raw data (CSV)" link with the per-model and combined numbers for that week. Updated 2026-W24.