How we rank: criteria and method
Six criteria, each weighted and scored 0–10 on a written scale. The overall score is the weighted mean. Every fact behind a score comes from the vendor's own pages, with the date it was checked.
The criteria and their weights
| Criterion | The question it answers | Weight |
|---|---|---|
| Citation tracking coverage | Does it tell you when AI answers name your brand and cite your pages, and how often? | 25% |
| Answer engines covered | Which AI answer engines does it check, and are they in the base plan? | 15% |
| API and MCP access | Can your own code and AI agents read and act on the data? | 15% |
| Pricing transparency | Can you see what it costs before talking to sales? | 10% |
| Actionability | Does it help you change the result, not only report it? | 20% |
| Evidence quality | Can you see the answers and sources behind each number? | 15% |
Why these weights: tracking whether AI answers cite you is the job, so it weighs most. Changing the result comes next, because a report nobody acts on changes nothing. Engines, access for your own code and agents, and evidence matter equally. Pricing transparency weighs least, but it decides whether you can try a tool this week.
Citation tracking coverage
Does it tell you when AI answers name your brand and cite your pages, and how often?
What is measured
Whether the tool tracks brand mentions, cited URLs and share of voice across a prompt set you control, and whether it also sees the other side: AI crawlers reading your site and visits that arrive from AI answers.
How to re-check it
Open the vendor's features or docs page and look for prompt tracking, cited-source lists, share of voice and crawler or referral analytics.
The scale
- 9–10
- Mentions, cited URLs and share of voice on your own prompts, plus AI crawler and AI-referral data from your site.
- 7–8
- Mentions, citations and share of voice on your own prompts and competitors.
- 5–6
- Mentions and citations, with limited prompt control or competitor views.
- 0–4
- Mentions only, or a sample of prompts you cannot choose.
Answer engines covered
Which AI answer engines does it check, and are they in the base plan?
What is measured
The number of answer engines tracked (ChatGPT, Perplexity, Google AI Overviews and AI Mode, Gemini, Claude, Copilot and others) and whether they come with the plan or cost extra.
How to re-check it
Read the vendor's pricing page: engine lists and add-ons are usually stated per plan.
The scale
- 9–10
- Six or more engines, the main ones included in the base plan.
- 7–8
- Four or five engines, or more with several as paid add-ons.
- 5–6
- Two or three engines.
- 0–4
- One engine, or the list is not published.
API and MCP access
Can your own code and AI agents read and act on the data?
What is measured
A documented public API, an official MCP server that agents such as Claude Code or Cursor can call, exports, and whether these are on every plan.
How to re-check it
Look for API docs and an MCP page on the vendor's site; check which plans include them.
The scale
- 9–10
- Documented API and official MCP server on every plan, able to read and act.
- 7–8
- Documented API and MCP server, on some plans or read-only.
- 5–6
- API or MCP, not both, or on enterprise plans only.
- 0–4
- CSV export or nothing published.
Pricing transparency
Can you see what it costs before talking to sales?
What is measured
Whether prices, plan limits and overage are published, and whether you can start without a sales call.
How to re-check it
Open the vendor's pricing page without signing in.
The scale
- 9–10
- Every plan priced, limits and overage published, self-serve start.
- 7–8
- Entry plans priced and self-serve; top tier on request.
- 5–6
- A starting price only, or prices behind a sign-up.
- 0–4
- Contact sales for any price.
Actionability
Does it help you change the result, not only report it?
What is measured
Whether the tool turns findings into work: specific recommendations, content it drafts, fixes it makes to your site, or changes it ships for review.
How to re-check it
Look for recommendation, content or site-fix features on the vendor's product pages.
The scale
- 9–10
- Makes the change itself (content, site fixes) for you to review and ship.
- 7–8
- Specific recommendations or drafted content tied to the gaps it found.
- 5–6
- General recommendations.
- 0–4
- Reports only.
Evidence quality
Can you see the answers and sources behind each number?
What is measured
Whether every metric links back to the raw AI answers and cited URLs it was computed from, with dates, so a number can be checked rather than trusted.
How to re-check it
In a trial, click through any metric and see whether it opens the answers and sources it came from.
The scale
- 9–10
- Raw answers, cited URLs and dates behind every metric, exportable.
- 7–8
- Raw answers and cited sources visible in the app.
- 5–6
- Cited sources shown, answers summarised.
- 0–4
- Scores without the answers behind them.
From six scores to one
Overall = the sum of each criterion score times its weight, divided by 100, rounded to one decimal. Ties go to the tool with the higher citation tracking score.
Worked example, Letter: 9 × 25 + 4 × 15 + 10 × 15 + 10 × 10 + 10 × 20 + 7 × 15 = 840, divided by 100 = 8.4.
The rules we keep
Only what is published
Facts come from each vendor's official pricing, feature and docs pages. Where a fact isn't published, the review says "not verified" and scores what is stated, never a guess.
No invented numbers
No ratings, reviews, prices or statistics we can't point to. Prices are quoted as the vendor's page showed them, in the currency it showed.
Dated
Every review shows the date its facts were checked. If a vendor's page now says something else, the page wins until the next review.
Same scale for everyone
Every tool, including the one ranked first, is scored on the same six criteria and the same written scale.
How to re-check the ranking yourself
- Open a review on the reviews page and note the sources and the date checked.
- Open each source. Look for what the criterion's "how to re-check it" line names.
- Score the criterion against the written scale above.
- Recompute the overall score with the weights, or download the data as JSON and do it in a sheet.