See who wins the answer, and open the evidence behind it.
When an AI answer names three companies and you are not one of them, the useful question is not how far behind you are. It is what the answer was built from, and which of those inputs you can actually change.
- Share of voice
- Every brand named across your valid checks, not just the rivals you already track.
- Evidence-linked
- Every gap opens onto the specific prompts, answers and citations behind it.
- Language rule
- Findings are stated as association. Causation claims are not permitted anywhere in the product.
Your slice of the answers in your category.
Share of voice is computed over valid checks only: every brand the answers named, divided by the answers we were actually able to observe. Because the denominator excludes provider failures, a bad API day cannot make a competitor look ascendant.
Share of voice · 240 valid checks · 5 engines
- Northwind31%
- Acme24%
- Initech17%
- All other brands28%
- Northwind19%
- Acme11%
- Initech8%
Presence and endorsement are tracked apart
- A brand can hold a large share of mentions while almost never being the one the answer actually recommends. Collapsing the two hides that.
- Competitors are discovered from the answers themselves, so the list includes rivals your positioning deck never mentioned.
- You can pin the brands you care about, and everything else rolls into an “all other brands” bucket rather than disappearing.
- Share of voice is available per engine, per market and per intent class, because a category-level average hides where the loss is concentrated.
Being named is not the same as being recommended.
Every appearance is classified into one of four types before it reaches a metric. The distinction between recommended and mentioned is the distinction between winning the answer and appearing as a footnote inside someone else’s win.
| Type | What it means | How it is scored |
|---|---|---|
| Recommended | The answer actively puts the brand forward as a choice for the question asked. | Feeds both the visibility and the recommendation component. |
| Listed | The brand appears inside a set or ordered list of options without a distinguishing endorsement. | Counts as presence, and carries a position where the list is ordered. |
| Mentioned | The brand appears in passing — an aside, a historical note, a comparison footnote. | Counts as presence only. It does not earn recommendation credit. |
| Absent | The engine answered and the brand was not there at all. A genuine, countable miss. | A valid observation that lowers visibility. |
An explanation you can argue with.
Each gap is grouped into a pattern and attached to the observations that produced it. The wording is deliberate: these are associations drawn from what we observed, and every one of them opens onto the underlying answers.
- Evidence suggests
Citation gaps
17 answers cite a source that names Northwind and never names Acme
The sources feeding these answers describe a shortlist you are not on. Closing that gap is a content and earned-placement problem, not a keyword problem.
- Likely contributing factor
Missing comparison coverage
6 comparison prompts with no owned page addressing the topic
When a buyer asks the engine to compare, the answer is assembled from whatever comparison material exists. Yours is not in the pool.
- Associated with
High-intent prompt losses
12 transactional and comparison prompts lost across 4 engines
Losses concentrated at the decision point cost more than losses on definitional questions, and intent weighting reflects that in the score.
- Evidence suggests
Ambiguous entity description
9 answers describing Acme in a category adjacent to the one you sell into
If the engine has learned an imprecise description of what you are, it will surface you for the wrong questions and skip the right ones.
The language rules behind these cards
Permitted phrasing: “evidence suggests”, “associated with”, “likely contributing factor”. Forbidden: causal claims such as “this caused” or “this is why you rank below”, and any promise about a future outcome.
The rule is enforced by a lint test over the explanation templates, so a future copy edit cannot quietly promote an association into a causal claim.
Open one competitor and read their whole footprint.
A competitor row expands into the full record: where they appear, how they are characterised, which prompts they take from you, and which sources keep putting them in front of the engine.
Northwind
Discovered from answers- Share of voice
- 31%
- Recommended in
- 19%
- Prompts won vs you
- 38
- Engines present on
- 5 of 5
- saasreviewhub.comReview platform
- northwind.com/compareCompetitor-owned
- devforum.ioCommunity
What the detail view answers
- Which specific prompts they take, sorted by intent weight so the expensive losses surface first.
- How the engines characterise them — the sentiment distribution of the language used around the brand.
- Which engines they are strong on and which they are thin on, since the gap is rarely uniform.
- Which cited sources their presence travels with, so an earned-placement plan has a target list rather than a vibe.
- How their share has moved across your scan history, with the sample size for each period.
Competitors reach the list two ways: extracted from the answers we observed, or added by you. Brand extraction that cannot be resolved confidently is left unresolved rather than guessed into a name.
The gap is not evenly spread across question types.
Split by intent, the picture usually changes. A brand can lead on definitional questions and still lose every question a buyer asks at the point of decision — which is the split that decides revenue.
Acme vs Northwind · visibility by intent
| Intent | Acme | Northwind | Prompts | Reading |
|---|---|---|---|---|
| Transactional | 18% | 39% | 22 | Pricing and purchase questions. The widest gap in the set. |
| Comparison | 21% | 44% | 31 | Versus and alternative-to questions where a shortlist is formed. |
| Commercial | 29% | 33% | 26 | “Best tool for …” evaluation questions. Roughly even. |
| Informational | 36% | 25% | 19 | Definitional questions, where your documentation is doing well. |
Every cell opens onto the underlying checks, so a claim on this table can be traced to the individual answers that produced it.
Find out who AI recommends instead of you.
One free check across the answer engines your buyers actually use.
No signup required for your first visibility check.
Every figure AnswerRank reports carries its sample size and how it was collected.
Read the methodology