On this page
Understanding results
Understand what your numbers mean, check the evidence behind them, and make fair comparisons over time.
Earlier Perplexity results
Perplexity tracking is temporarily unavailable. Existing projects continue checking their other selected platforms. Your saved Perplexity answers, citations, and exports remain available in Saved results. New tracking selections require at least three available platforms; an existing selection reduced by this interruption can continue without adding a replacement. Perplexity referral traffic and crawler activity are separate features and remain available.
Bing AI Mode and earlier Copilot results
Bing AI Mode tracks answers from Copilot Search in Bing. Keep each prompt focused on a single question; very long prompts and conversations with follow-up turns are skipped for this platform. Earlier Microsoft Copilot answers keep their original platform name and remain available in Saved results. The two platforms have separate histories and comparison baselines.
In a Bing answer, citation groups show the pages linked together at a point in the answer. A page can appear in several groups. All sources also includes pages read for the answer that were not cited inline. Shopping cards and sponsored placements are unavailable for this platform.
Start with the scope
Ansvra describes the saved answers included in your current view. It does not measure every AI conversation or the total demand for a question. Before reading a number, check which answers it represents.
Check the result view
Use Current results for your project’s current view, or Result history to find Earlier results. Read the selected dates before comparing results.
Check the filters
Review the date range, country, language, platforms, and prompts or topics where those filters are available. Different views can include different sets of answers.
Read the answer count
Look for the number of usable answers behind the metric. A percentage based on a handful of answers gives you less evidence than the same percentage based on many answers.
- ChatGPT
- Google AI
- Gemini
- Bing AI Mode
Usable answers are saved answers available for measurement in that view. Failed, pending, or missing answers are not treated as usable answers in which your brand was absent.
Brand visibility: are you mentioned?
Brand visibility is the percentage of usable answers in the selected scope that name your brand, based on its configured names and aliases. Responses naming you is the underlying answer count.
In the preview, Alder & Coast appears in 175 of 432 answers. Dividing 175 by 432 and multiplying by 100 gives approximately 40.5% visibility. Naming the brand several times in one answer still contributes one answer to this count.
Website citations: is your site referenced?
Website citations, Website citation rate, or Answers citing you describe how often usable answers reference a page on your configured website domain. Read the label beside the number: a percentage of answers is different from a count of cited pages.
The answer-based rate is answers citing your site ÷ usable answers × 100. In the preview, 175 of 432 answers cite Alder & Coast’s website. That is about 40.5%; a view rounded to whole percentages shows 41%.
A brand mention and a website citation are separate signals. An answer can name you without citing your site, or cite your site without explicitly naming your brand.
- Citation occurrences
- Cited-page appearances across the selected usable answers. The same page can contribute again when it is cited in another answer; repeated references to the same normalized page within one answer count once.
- Unique cited pages
- The distinct cited pages in the selected scope. A page cited by several answers still contributes one unique page.
- Unique domains
- The distinct website domains represented by those cited pages. Several pages can belong to one domain.
Read results for an individual question
Open Tracked Prompts to see which questions contribute to your overall picture. A prompt’s Brand visibility describes its own eligible answers, so it can differ from the overall percentage.
Your site citations and All cited pages, when shown as prompt columns, count cited-page appearances across answers rather than percentages. Tracked-brand appearances counts each eligible tracked brand once per answer in which it appears. Competing brands identifies tracked competitors mentioned in the prompt’s results.
AdvancedWhy the overall rate is not the average of prompt percentages
An answer-based overall rate uses the total qualifying answers in that view. Prompts with more answers therefore contribute more to that rate. Averaging displayed prompt percentages gives every prompt the same weight and can produce a different result. Prompt metrics also match the prompt’s saved country and language; check targeting when comparing a prompt with broader reports.
Compare tracked brands fairly
- Share of voice
- Your brand’s answer appearances as a share of all appearances for the included tracked brands. An answer that names several tracked brands contributes one appearance to each of them.
- Mention order
- The average order in which your brand first appears relative to the other included tracked brands, calculated only for answers mentioning you. A lower number means your brand appeared earlier among that tracked set.
AdvancedTwo percentages with different denominators
Visibility divides your brand’s answer count by usable answers. Share of voice divides your brand’s answer count by the sum of answer appearances for the included tracked brands. For an illustrative comparison, if your brand appears in 20 of 100 usable answers, visibility is 20%. If the included tracked brands have 50 appearances in total, your share of voice is 40%. These illustrative figures are not the Alder & Coast screenshot’s measurements.
Understand additional signals
- Net brand tone
- When available, this summarizes classified brand mentions on a scale from −100 to +100. Positive mentions move it upward; negative mentions move it downward. Review the classification coverage and original wording before drawing conclusions.
- Ad appearance rate
- The share of eligible answers with known advertising information in which ads were present. It describes ads in the returned answer, not necessarily an ad for your brand, and does not measure advertising spend.
- Product-card rate
- The share of eligible answers with known product-card information in which shopping or product cards were present. It does not mean your own products appeared or that a purchase happened.
AdvancedHow net tone and its coverage work
Net brand tone is (positive classified mentions minus negative classified mentions) divided by all classified mentions, multiplied by 100. Neutral and mixed classifications remain in that denominator; unknown or unclassified mentions do not. Coverage describes how much of the eligible mention evidence has been classified. A score of zero can reflect neutral or mixed results, or a balance of positive and negative classifications. It does not establish universal neutrality.
Read changes and gaps carefully
For counts and appearance rates, a zero means eligible evidence was available and none met the measure’s condition. Net brand tone is different: zero can also mean positive and negative classifications balance out. A dash or an unavailable value can mean there is no suitable evidence or comparison. Read the nearby explanation before deciding what it means.
Dates without usable answers are gaps, not observations of zero visibility. On charts that bridge gaps with dashed lines, the line connects observed points; it does not fill in missing measurements.
AdvancedPercentage points and comparable evidence
A move from 40% visibility to 50% is an increase of 10 percentage points, not 10% relative growth. Comparable change calculations use matching evidence groups across the periods; a headline rate based on all selected answers can therefore differ from the rate in a comparable subset. If there is no matching earlier evidence, a change may be unavailable. Check the comparison’s scope instead of assuming two nearby percentages were calculated from identical sets.
Turn a number into a useful next step
Pick one relevant question
Choose a prompt that matters to your customers, rather than investigating every movement at once.
Open its saved answer
Read the answer alongside its date, platform, and market. Check what it says about your brand and which pages it references.
Separate the observation from your interpretation
“Our site was cited in this answer” is an observation. “This citation brought us new customers” needs separate evidence. Record a specific question or improvement to follow up on.
Why do two screens show different percentages?
Check their result view, dates, targeting, prompts, platforms, and rounding. A prompt-specific rate and an overall rate can both be correct for different answer sets.
Does no result mean my brand was absent?
Not by itself. A failed, pending, unavailable, or missing answer is not a usable answer with a confirmed absence. Read the collection or availability status before interpreting it.
Can I infer customer demand from my prompts?
No. Tracked prompts define the questions you measure. Suggested prompts and the number of saved answers are not measures of how frequently people ask those questions.