How Parse measures AI visibility
The data we collect and how we score it.
Where every Parse number starts
Parse observes how public AI assistants answer real buyer questions, then extracts the brands, citations, and sources they use.
Prompts
10.5K+organic AI-search queries indexed
Responses
3.1M+AI responses observed
Brands
688.2K+brands tracked
Citations
36.9M+citations analyzed
Where prompts come from
We index organic AI-search prompts, the real questions buyers ask AI assistants.
- Real questions buyers ask, not keyword lists
- Source: aggregated public AI-search behavior
Models we query
| Model | Coverage | Refresh |
|---|---|---|
| ChatGPT | All prompts | Daily |
| Google AI Overviews | All prompts | Daily |
What we extract from each response
- Brand mentions
- Every named brand that appears in the answer.
- Recommendation rank
- The order in which brands are listed when AI is asked who's best.
- Citations
- Outbound URLs the AI answer points to.
- Sources
- Root domains behind those citations.
What each number is measured over
Every figure on a Parse page has a denominator. This chapter names them, so a percentage on one page can be compared with a percentage on another.
The basis line, part by part
Public pages carry a line naming the evidence behind them. Here is one, glossed.
Based on 12,431 AI responses from ChatGPT and Google AI Overviews across 1,204 prompts
- Based on 12,431 AI responses
- Every individual answer Parse collected for this market in the period. One question asked on two platforms produces two responses, so responses usually outnumber prompts. This is the denominator for mention rate.
- from ChatGPT and Google AI Overviews
- The platforms behind those responses. Public pages read these two. A paid Monitor reads ChatGPT Search and Google AI Mode, which is a different pair of surfaces on the same two assistants.
- across 1,204 prompts
- The distinct buyer questions behind those responses. Evidence floors are checked against prompt count, not response count, so one question answered many times does not on its own clear a floor.
Which numbers are shares, and of what
| Number | Counts | Out of | Sums to 100% |
|---|---|---|---|
| Mention rate | Responses that name the brand | Every response collected for that market in the period | No |
| Share of voice | The same responses as mention rate, under an older name | Every response collected for that market in the period | No |
| Recommendation share | Recommendation slots the brand occupies | Every recommendation slot in that market | Yes |
| Parse Score | A composite of the three pillars | Not a share. An index from 0 to 100 | No |
Recommendation share is the only figure above that is a share of a fixed total. Mention rate and share of voice describe the same measured quantity, so several brands in one market can each be named in a large fraction of answers and the column will not add to 100%.
Collected daily, published weekly
Collection, publication, and delivery run on different clocks. The refresh column in chapter 01 reports the first of the three.
- Collected Daily
- Prompts are asked and responses are stored every day, which is what the refresh column above reports.
- Published Weekly
- Public rankings and market pages recompute on a weekly cycle, so a page can quote figures gathered over the preceding seven days.
- Delivered Weekly
- A paid Monitor sends one interpreted Reading per week. Collection underneath it still runs daily.
Which assistants a number came from
Parse runs two collection pipelines on two different pairs of surfaces, so platform names legitimately differ between the free index and a paid Monitor.
- Public pages read ChatGPT and Google AI Overviews. These are the two rows in the models table above.
- A paid Monitor reads ChatGPT Search and Google AI Mode.
Both pairs sit on the same two assistants. A figure from one pipeline is not directly comparable with a figure from the other, because the surfaces answer differently.
Evidence floors and precision
A comparison axis backed by fewer than 3 claims, or drawn from fewer than 2 distinct prompts, does not earn a verdict. It reports insufficient evidence instead. This is why a young market shows no 30-day change for a while: the evidence to support one has not accumulated yet.
- Parse Score and pillar values
- Shown to one decimal place, here and on brand pages. A score of 72 prints as 72.0.
- Head-to-head side scores
- Rounded to a whole number. The comparison card deliberately shows a coarser figure than a brand page, because far fewer claims sit behind it.
- Large counts on summary cards
- Abbreviated to one decimal in thousands or millions with a plus, for example 12.4K+. Exact counts appear in the basis line instead.
- Suppressed values
- A number that fails an evidence floor is not rounded to zero. It is not rendered at all.
Three pillars, one composite
Every brand earns a Parse Score from three independently measured pillars. Each pillar answers a different question.
Pillar
Strength
How prominently you rank.
"When AI mentions you, how prominently do you rank?"
Input
Where you rank in the answers where you appear.
Output
0 to 100Pillar
Reach
How broadly AI mentions you.
"Across how many topics does AI know you exist?"
Input
Prompts you appear in versus prompts where you'd be eligible.
Output
0 to 100Pillar
Authority
How well trusted sources back you.
"When AI shows its sources, how well are you backed?"
Input
Trusted third-party sources cited in answers that mention you.
Output
0 to 100The composite
The three pillars combine into a single Parse Score from 0 to 100. The composition method is held stable for at least a year before any revision, so the score remains a fair benchmark across time.
Pillar values are public on every brand page. Exact weights are held internal.
How to read a score
Parse Score uses neutral population bands from the rank-curve scale.
Calibration note: these bands describe the displayed Parse Score. Individual pillar values use magnitude color only; a Reach of 60 is not a population-rank claim.
How Stripe's Parse Score was built on Apr 24, 2026
A walkthrough using one brand on one day.
Input data
- 272prompts where Stripe was eligible to appear
- 217prompts where Stripe actually appeared (a 79.8% appearance rate)
- 12.4KAI responses analyzed across 90 days
- 78distinct sources cited in answers mentioning Stripe
- 2.1Kcitations from sources in answers mentioning Stripe
Pillar
Strength
When Stripe appears in an answer, AI tends to place it near the top. On average Stripe ranks near #2 across the prompts where it shows up.
Pillar
Reach
Stripe appears across most of the prompts where we'd expect a payment processor to show up. After accounting for niche concentration, Reach lands in the upper band.
Pillar
Authority
When AI cites sources in answers mentioning Stripe, trusted third-party sources appear often enough to give Stripe a strong Authority signal.
Composite
Parse Score
The three pillars combine into Stripe's composite Parse Score of 83.0.
How a head-to-head verdict is decided
Comparison pages break a matchup into axes, then grade the evidence behind each one. The grade decides what the page is willing to say.
Which way an axis leans
Each axis resolves to one of four leans. The page prints the lean as a short tag beside the axis name.
- Leans to the first brand
- renders as The first brand's name
- Leans to the second brand
- renders as The second brand's name
- Even
- renders as Even
- Mixed
- renders as Mixed
An axis that falls under the evidence floor does not receive a lean. It reports insufficient evidence.
How strong the agreement is
Consensus is graded into four levels. Confidence appears only as a flag on the exceptions: a solid verdict renders nothing at all, and only a disputed or thinly evidenced one is labelled.
At least 6 claims across at least 2 prompts, the leading side holds 70% or more of them, and the opposing side holds 25% or less.
At least 4 claims across at least 2 prompts, and the leading side holds 55% or more of them.
Both sides carry at least 2 claims, or the gap between the leading and opposing sides is 15 percentage points or less.
Everything else, including any axis under the evidence floor of 3 claims and 2 prompts.
Reading the absence: no confidence flag beside an axis means the evidence was strong or moderate. Absence is the signal, not an omission.
What a comparison counts
A comparison is measured over claims, not responses. A claim is one statement an assistant made about one of the two brands on one axis. The basis line on a comparison page therefore counts claims, while a market page counts responses. The two are not interchangeable.
Side scores on a comparison card are rounded to whole numbers, which is coarser than the one decimal a brand page shows. The claim counts behind them are small enough that a decimal would imply precision the evidence does not carry.
Core metrics
A reference index of every metric that appears in the product, with a one-line definition for each.
Brand metrics
Metrics that measure a brand's AI visibility and performance in AI answers.
Parse Score
Your AI-visibility rank
Your AI-visibility rank among tracked brands, mapped to a 0-100 score.
More detail
A single score that summarizes how prominently you appear, how broadly AI mentions you, and how well trusted sources back you.
Metric type: Score (0-100)
Strength
How prominently you rank
How prominently you rank when AI mentions you, adjusted for data confidence.
More detail
When AI includes your brand in an answer, Strength measures how prominently it places you and adjusts that signal for data confidence.
Metric type: Score (0-100)
Reach
How broadly AI mentions you
How often your brand appears across AI responses.
Metric type: Score (0-100)
Model visibility
How often the brand appears in ChatGPT and Google AI Overview responses.
More detail
Combined score showing the brand's presence across major AI platforms. Calculated from visibility rates in both ChatGPT and Google AI Overview.
Metric type: Score (0-100)
Peer visibility
Competitive standing versus tracked peers in your category.
More detail
Shows how the brand's visibility compares to direct competitors and similar brands in the same industry. This highlights competitive position in AI responses.
Metric type: Score (0-100)
Mention rate
Percentage of analyzed AI answers that name this brand.
More detail
The share of analyzed AI answers where the brand is named, over a defined set of questions: a market's buyer questions on a ranking page, one exact question phrasing inside the product. It is a rate, not a split — every brand in a market can be named often, and the rates do not sum to 100%. A brand at 15% is named in roughly 15 of every 100 answers. For pooled estimates across a question group's phrasings, see Visibility.
Metric type: Rate (percentage)
Average position
Average ranking position when the brand appears in AI responses.
More detail
When the brand is mentioned, this shows where it typically appears in the response (1st, 2nd, 3rd, etc.). Lower numbers mean more prominent placement.
Metric type: Rank (lower is better)
Parse rank
Ranking position among all brands in the index.
More detail
Where the brand sits compared to every other brand that we track. A lower rank means it appears more frequently across the full index.
Metric type: Rank (lower is better)
Rank change (30d)
How many positions the brand moved up or down in the past 30 days.
More detail
Compares the brand's current rank to its rank 30 days ago. Useful for spotting longer-term trends beyond weekly noise.
Metric type: Delta (signed number)
30-day trend
Sparkline showing the brand's visibility trajectory over the past 30 days.
More detail
A miniature chart of daily visibility scores. An upward line means the brand is becoming more visible to AI over time.
Metric type: Score (0-100)
Mentions
Number of AI responses that include this brand during the selected period.
More detail
Raw count of how often the brand shows up in AI answers. Track this alongside visibility and rank metrics to spot meaningful volume changes.
Metric type: Count
Citation metrics
Metrics about how AI cites and references sources.
Trust
How strongly AI relies on this domain
0-100 source-domain reliability score based on citation volume, prompt breadth, and cross-platform consensus.
More detail
Trust scores how broadly and consistently AI relies on this domain as a citation source. It is maintained from recent citation-domain daily metrics: citation volume carries the most weight, prompt breadth next, and cross-platform consensus rounds out the score.
Metric type: Score (0-100)
Citations
Number of times AI models cite the domain.
More detail
Counts how often AI responses attribute information to the domain. More citations signal that models trust and rely on the content.
Metric type: Count
Pages cited
Total number of pages from this domain referenced by AI models.
More detail
Measures how many individual pages from the domain were cited over the selected period. Useful for spotting domains that provide deep coverage on relevant topics.
Metric type: Count
Unique URLs
Number of unique URLs from the domain cited by AI models.
More detail
The breadth of pages from the domain that AI models reference. A higher count shows that multiple pieces of the content are cited, not just a single flagship page.
Metric type: Count
Prompts
Unique prompts that cite or reference this domain.
More detail
Counts how many distinct prompts include this citation across AI responses. Higher values mean the domain appears across a broader set of user intents.
Metric type: Count
Last cited
Most recent time AI models referenced this domain.
More detail
Shows how fresh the citation is. Recent timestamps signal that the domain is still being referenced frequently.
Metric type: Date
Prompt metrics
Metrics about brand performance within specific AI prompts.
Prompt visibility
0-100 score based on mention rate and position ranking.
More detail
Overall prompt visibility on a 0-100 scale, based on how often your brand appears and how prominently it is shown.
Metric type: Score (0-100)
Total brands
Number of unique brands mentioned in responses to this prompt.
More detail
The diversity of brands that models mention for this prompt. A higher number means answers are pulling from a broader competitive set.
Metric type: Count
Brand mentions
How many times this brand is mentioned in responses to the prompt.
More detail
Total mention count for the brand when people ask this prompt. Pair it with visibility to understand both frequency and reach.
Metric type: Count
Mentioned
Whether the focus brand appears in an individual response.
More detail
Shows if the selected brand is referenced in a specific AI response so you can quickly zero in on answers that include you.
Metric type: Boolean
Change
Change in metric value over the selected time period.
More detail
Represents the delta between the current value and the prior comparable period. Positive change shows improvement, negative values indicate decline.
Metric type: Delta (signed number)
Perception metrics
Metrics about the language and tone AI uses when describing a brand.
Sentiment mix
Share of positive, neutral, and critical or mixed language in evidence about the brand.
More detail
Sentiment mix groups the evidence language around a brand into positive, neutral, and Critical or mixed buckets so framing changes are visible without hiding ambiguous criticism.
Metric type: Rate (percentage)
Descriptor frequency
How often a specific descriptor appears in evidence about the brand.
More detail
Descriptor frequency counts recurring words and phrases used to describe a brand in AI evidence, helping show which associations are becoming more or less common.
Metric type: Count
Descriptor polarity
Whether recurring descriptors around the brand are mostly positive, neutral, or critical or mixed.
More detail
Descriptor polarity classifies recurring descriptors by tone so a brand can separate harmless volume shifts from changes in how AI frames the brand.
Metric type: Score (0-100)
Sector metrics
Metrics that compare a brand within a selected public-index sector.
Sector Score
A sector-local visibility score for brands within the selected sector.
More detail
Sector Score ranks brands within one sector by combining sector-local strength and reach. It is scoped to the selected sector and is separate from the index-wide Parse Score.
Metric type: Score (0-100)
Sector Strength
How strongly the brand performs in prompts tied to the selected sector.
More detail
Sector Strength measures the brand's sector-local visibility quality across prompts assigned to the selected sector.
Metric type: Score (0-100)
Sector Reach
How broadly the brand appears across markets in the selected sector.
More detail
Sector Reach measures how broadly the brand appears across the selected sector's markets, normalized within that sector.
Metric type: Score (0-100)
Global Parse Score
The brand's index-wide canonical Parse Score, shown beside sector-local scores for context.
More detail
Global Parse Score is the same canonical Parse Score used across the public index. Sector pages show it beside Sector Score so sector-local strength can be compared with the brand's overall AI visibility.
Metric type: Score (0-100)
