Methodology
How We Score Tools
We publish the rubric because a score you cannot check is just an opinion with a number attached.
We do not sell any tool we score. We do not accept payment for placement, ranking position, or inclusion. Where a scored tool publishes its own competing ranking, we say so in the entry.
The rubric
Every tool is scored out of 100 across six criteria.
| Criterion | Weight | What earns points |
|---|---|---|
| Cost efficiency at entry tier | 30 | What the cheapest real plan costs relative to what it includes |
| Engine coverage at entry tier | 25 | How many AI engines the entry plan actually tracks |
| Pricing transparency | 15 | Full published pricing scores 15, partial scores 5 to 7, quote-only scores 0 |
| Capability depth | 15 | Sentiment, citations, competitor benchmarking, API, MCP, distinctive diagnostics |
| Action layer | 10 | Whether it only reports the gap or helps close it |
| Free tier or trial | 5 | How cheaply you can evaluate it before committing |
Four decisions that shape the numbers
We score the entry tier, not the maximum tier.
A platform may support ten engines while the plan you can afford tracks one. Scoring maximum coverage rewards exactly the practice that catches buyers out. If a vendor gates engines behind higher pricing, that shows up as a lower score.
We do not exclude vendors who hide their pricing.
Several published rankings drop quote-only vendors for having no public pricing page, which conveniently removes competitors. We keep them in and score them zero on transparency. A buyer can decide whether opacity matters to them.
We use bands, not decimals.
Cost per prompt per engine is the metric that matters, but prompt allowances are inconsistently published and credit-based vendors convert credits to prompts differently by engine and mode. Publishing a figure like $1.87 per prompt per engine across the whole category would be false precision. We band instead, and say so.
We publish the arithmetic.
Every scored entry shows its component scores, not just the total. If you disagree with a weighting, you can recompute with your own.
What this score does not measure
This matters more than the rubric itself.
The score measures price, disclosed capability, and how much of the product you get at the entry tier. It does not measure:
- •Output accuracy, or whether the data a tool reports is correct
- •Quality of support
- •Whether the interface is pleasant to use
- •Reliability, uptime, or how the product performs under load
- •How good the vendor's recommendations are
A low score means expensive, gated, or opaque at the entry tier. It does not mean bad. Several of the lowest-scoring tools in our AI visibility ranking are the strongest products in the category for enterprise buyers, and the rubric is close to useless for that buyer. Read the use-case sections rather than the ranking if you have real budget.
What we have not done
We have not independently trialled every tool we score. Where a score rests on vendor-published or third-party reported figures rather than our own testing, the entry says so.
This is a real limitation and we would rather name it than imply testing we have not done. As we complete hands-on evaluations we will add a tested-accuracy criterion and reweight, and the change will be dated on this page.
Freshness
Pricing in these categories changes monthly. Every scored ranking carries the date its figures were checked. Scores older than one quarter should be treated as indicative rather than current.
Where sources conflict on a vendor's pricing, we say so in the entry rather than picking one and presenting it as settled.
Corrections
If a score is wrong, or a vendor's pricing has changed, tell us and we will update it and note the change. Corrections are made to the page rather than quietly overwritten.