On this page
Your product has a full name, a shorter name and an abbreviation. Adding all three to monitoring seems sensible: you want to recognize the ways people refer to the same brand.
The difficulty is that a shorter name may also describe something else. A shared word, an abbreviation or the name of a related product can turn an apparent appearance into an identity question.
A useful alias identifies the same entity under a legitimate alternative name. Before accepting it, inspect the actual answer context and decide what evidence would distinguish your brand from another meaning.
Define what you are measuring first
Decide whether the target is the company, a particular product or a product family. They may be related without being interchangeable for your report.
If a company sells several products, a mention of the parent company does not automatically establish that an answer recommended the particular product being studied. Equally, an old company name may be relevant to historical analysis without being the right default for a current product-discovery report.
Write the intended entity and approved domain beside the monitoring definition. Keep aliases separate from competitors and from the questions used to collect answers. Changing question wording changes what you ask; changing an alias changes how an answer may be interpreted.
The prompt-selection guide covers which questions to track. This review focuses on which entity a recorded name refers to.
Build a candidate list from real usage
Start with names supported by actual product materials or clear observed references: an official spacing variant, a documented former name, a recognized localized form or a product abbreviation used in relevant context.
Treat these as candidates to inspect, not a list that becomes better as it gets longer. Do not add a broad category term simply because it appears in many answers about your market.
Illustrative example only. HarborDesk is a fictional desk-booking product used to explain the review. The names and passages below are not observations about real companies.
| Candidate | Why it might help | What needs checking |
|---|---|---|
| HarborDesk | Primary product name | Whether the answer refers to the intended product rather than a namesake |
| Harbor Desk | Possible spacing variant | Whether legitimate usage connects it to the same product |
| HD | Possible abbreviation | Whether the surrounding answer clearly establishes that meaning |
| Harbor | Shorter shared word | Whether it identifies the product or something unrelated |
| Desk booking | Product category | It describes a kind of solution and should not be accepted as a brand alias merely for that reason |
An approved domain can help resolve identity when a relevant link is present. Its absence is not automatic proof that the name is wrong, and a link to a third-party page does not make that page an owned source.
Inspect what the passage refers to
Read the sentence around the match, the question being answered and any relevant links. Look for the product's function, organization or other identifying context.
The phrase “harbor desk booking” could refer to a service at a marina. In an answer that first spells out the fictional product name and then says “HD,” the abbreviation has a much clearer referent. A matching word alone does not settle either case.
This is a review of meaning, not a claim that every monitoring system uses literal substring matching. Detection and classification methods differ. The practical question is whether the recorded result identifies the entity your report intends to measure.
Also keep identity separate from recommendation. “HarborDesk does not support the requested feature” identifies the brand, but does not recommend it for that need. A correctly identified mention can still be negative, neutral or a reason to exclude a product.
Use the mentions, citations and recommendations guide to keep those observations distinct.
Check questionable matches and missed appearances
Reviewing only the matches that look positive can miss two different problems: counting an unrelated entity and failing to recognize the intended one.
Include clear full-name appearances, ambiguous short names, apparently unrelated matches and suspected missed references. Keep unresolved cases visible rather than silently treating them as confirmed appearances or confirmed misses.
For an illustrative review of ten successful answers, suppose a loose candidate rule flags six as possible appearances. Reading those six finds four clear references to the intended product, one unrelated use of “Harbor” and one ambiguous abbreviation. The other four answers have no observed target reference under the review method.
A transparent note says: “Four confirmed appearances, one unresolved candidate and one rejected false match across ten successful answers.” A conservative confirmed-appearance rate would be 4/10, with the unresolved case disclosed separately. Calling all six confirmed appearances would overstate the evidence.
That is a manually defined review calculation. It is not a claim that PB automatically produces these adjudication categories or this adjusted percentage. Even a corrected count remains bounded by what the review examined.
Do not count several aliases as several appearances
An answer might say “HarborDesk, also written Harbor Desk” and later use “HD.” Those three strings may identify one product in one response.
For a response-based appearance rate, count that response once. Do not add separate appearances for every occurrence or alias. An occurrence count would be a different metric and would need its own definition.
PB's AI Visibility overview uses the configured brand and aliases when interpreting recorded target mentions, and a successful visible response contributes once to its Visibility numerator. That still leaves the identity judgment worth checking when an alias is ambiguous.
Keep competitor aliases attached to the appropriate competitor. PB's competitor workflow provides an Other names field for alternative names; adding a category term there can create the same interpretation problem as adding an overbroad target alias.
Record a definition change before interpreting a trend
Adding, removing or correcting an alias can change the classification of available evidence. A different count after a definition change does not, by itself, show that a provider produced different answers.
Keep a short manual change record:
| Field | What to record |
|---|---|
| Entity and definition version | The company/product and the alias set used |
| Change and date | Candidate added, removed or corrected, with the review time |
| Reason | The actual passage or product evidence supporting the decision |
| Historical treatment | Which saved answers, if any, were reviewed again under the revised definition |
| Comparison limit | Which earlier figures still use the previous method |
If retained answers and available tools allow a consistent historical review, preserve both the original and revised interpretations with their methods. Otherwise, mark the definition break and use a new comparable baseline. Do not quietly splice old-method and new-method percentages into a seamless trend.
PB's competitor evidence can be revisited using current definitions, but do not assume every target summary and historical chart is rebuilt identically after any alias change. Confirm the affected view and preserve the original answer before deciding what changed.
Make the next review about the entity
When a result looks wrong, start with one saved answer, the configured name and aliases, and the specific phrase being counted. Ask whether it refers to the intended entity, what kind of appearance it is, and whether the same rule was used in the comparison period.
If the evidence and classification disagree, pass that exact record to the relevant reviewer or support workflow. Include any uncertainty. Adding more names until the percentage rises is not a reliable way to resolve the discrepancy.
For client reporting, put material alias changes in the methodology section of the AI visibility report. That keeps a clearer identity definition from being mistaken for a change in the answers themselves.