Brand measurement in 2027: a role-by-role guide to brand tracking, AI visibility and social listening
Brand trackers, social listening and AI visibility tools can rank the same competitors in a different order, and when different teams buy them against different competitor lists, the results cannot be read together. Our study of beauty brands in five markets shows where the sources differ, and each role in the marketing organisation has its own part in turning them into one view of the market in 2027.
Survey. Odyssey Beauty Signal in the UK, Sweden, Norway, Finland and Denmark. Online panel interviews with people aged 16 to 65 who had bought beauty products in the past six months, collected every month and weighted. 23,745 respondents in the 12 months to August 2026.
Brand measurement in 2027 works when every source is read against one view of the market, and getting there is an organisational decision more than a technical one. That view is set by segmentation, positioning and tracking, and they hold together only when they share one sample, one set of customer needs and one competitor list. Run as separate projects, they cannot be joined afterwards. A tracker, a social listening tool and an AI visibility tool can rank the same competitors in a different order, and without that view the company cannot tell whether the brand is doing differently or the measures are describing different things.
In our report The new rules of brand measurement 2027, we compared the three sources for the same beauty brands in the UK, Sweden, Norway, Finland and Denmark. The survey covers 23,745 beauty buyers over 12 months, the conversation data comes from TikTok, YouTube, podcasts and broadcast, and five AI search engines answered 60 product and 60 retailer questions per market in September and October 2026. The data comes from beauty, but the findings concern how measurement is set up, so they apply in other categories as well.
Who this guide is for
The guide is for everyone who works with the brand, whether in strategy, insights and communication or in the places where the brand is seen, from PR and social media to AI answers. Go straight to your role:
CMOs and marketing leaders, who set the direction for the brand and decide what its progress is measured against
Insights, who follow buyers, competitors and the market, and run the brand tracking
Brand and strategy, who decide what the brand should stand for and for whom
Procurement, finance and data teams play a part as well, covered under roles you may also need.
What we found in five markets
Behind every finding is one condition
None of these comparisons can be made unless segmentation, positioning, tracking, conversation and AI answers are measured against the same competitor set and the same customer needs. In our study the survey places every buyer in a demand space, a group of buyers who shop for the same main need, such as trust or affordability. It also places every brand on a positioning map, and conversation and AI answers were counted for the same brands in the same categories. Without that shared base, a brand's AI rank cannot be read against the need its buyers link it with, and its share of conversation cannot be set against its share of buyers.
Matilda's view
The sentence in this report I would underline is the one about projects being competently executed and still leaving the company without a consistent account of the market. I have sat in those reviews. The segmentation supplier, the tracker, the social listening tool and the AI visibility vendor each presented sound work, and none of it could be laid over the rest because the competitors, the needs and the periods were all different.
Matilda Rydow, CCO, Odyssey
1. An overall AI ranking can hide how a brand stands on the need its buyers link it with
We compared each brand's rank on all the questions in its category with its rank on the questions about the need its buyers associate it with. Of 47 of the most preferred brands, 11 moved three places or more or were not named at all for that need. Nivea falls from second to tenth in UK skin and body care, while Dermalogica rises from 13th to seventh among Norwegian skin and body care brands.
Five markets, the most preferred brands in each category where the need has at least three questions. AI answers from five engines, September and October 2026. Source: Odyssey.
Matilda's view
An AI visibility ranking tells me very little until I understand which consumer needs the questions represent. I want to know when a brand is named, for what need and against which competitors, and how that compares with the position it holds among buyers. You can't get that from AI data alone. You need brand positioning research in the category.
Matilda Rydow, CCO, Odyssey
2. Buyers and AI answers can favour different brands for the same need
Among UK skin and body care buyers who value trust most, 22% prefer Nivea and 6% CeraVe. In the AI answers to the trust questions, CeraVe has 64% of the tracked brands' mentions and Nivea 1%.
UK skin and body care. Survey: preferred brand among buyers in the trust demand space, 12 months to August 2026. AI: share of the tracked brands' mentions in answers to the trust questions, September and October 2026. Source: Odyssey.
Matilda's view
The differences between the sources are where the conclusions are. The survey tells me where buyers have the brand today. Organic mentions and AI answers tell me where the published conversation has it, which is often where the last few years of content and reviews have left it.
Matilda Rydow, CCO, Odyssey
3. Reading the answers by need shows where a brand's rank comes from
Trust questions make up 42% of the UK skin and body care questions, and 5% of Nivea's mentions come from them. Nivea's second place overall is built on other needs, such as quality, where it also ranks second, while on trust, the need its buyers associate it with most, it ranks tenth.
UK skin and body care, AI answers from five engines, September and October 2026. Source: Odyssey.
Fredrik's view
I don't think we should stop measuring overall AI visibility. There is value in understanding a brand's presence across the market. But we need to know what that overall figure represents.
Fredrik Augustsson, Chief Data Officer and Co-founder, Odyssey
4. AI mentions rank brands more like unaided awareness and consideration than like purchase
This held in all 20 combinations of market and product category, with unaided awareness closest in 13 and consideration in 7. In UK skin and body care, CeraVe has about 5% of the tracked brands' purchases and 35% of their AI mentions, while Dove has 13% of purchases and 3% of mentions.
UK skin and body care, 56 tracked brands. Survey: share of purchases, 12 months to August 2026. AI: share of mentions, September and October 2026. Source: Odyssey.
Fredrik's view
The link with unaided awareness makes sense mechanically. Spontaneous recall and an AI answer both start from a blank page; in both cases the brand has to come up without a list to prompt it. Aided awareness is a different kind of memory, built over years of shelf presence and advertising, which is why it can stay high for a brand the engines have stopped mentioning.
Fredrik Augustsson, Chief Data Officer and Co-founder, Odyssey
5. First places go to a few brands
In UK skin and body care, the five brands named first most often take 70% of the first places among the tracked brands. The five largest brands take 43% of purchases, and the five best known 18% of aided awareness.
UK skin and body care, 56 tracked brands. Survey: 12 months to August 2026. AI: five engines, September and October 2026. Source: Odyssey.
Fredrik's view
Mentions and first places are different measures and I would never let them share an axis. An answer that lists eight brands gives each of them a mention; only one is named first, and the engines hand that out far more narrowly than consumers hand out awareness. For a smaller brand that is the number to watch, because a respectable mention share can hide the fact that it is never the answer.
Fredrik Augustsson, Chief Data Officer and Co-founder, Odyssey
6. Three in four UK answers that name a brand start with one outside our 75-brand survey list
That is 964 of 1,290 answers, against a list already longer than the 20 to 30 brands a standard tracker covers. La Roche-Posay, on none of the five markets' lists, is named first more often than any brand except CeraVe. Dove, the second most purchased brand in UK skin and body care, is named first in six of 311 category answers, so a list built from AI answers alone would miss it.
UK AI answers that name a brand, five engines, September and October 2026. The list is Odyssey's UK survey list of 75 product brands. Source: Odyssey.
Matilda's view
Market research has been expensive and traditional, so companies have had to choose which competitors to track. Most chose the ones that were big at the time, and that leaves out the challengers that move fast. If the list stays the same year after year, a new challenger can grow into a serious competitor without the tracker ever asking about it.
Matilda Rydow, CCO, Odyssey
7. A first page on Google does not secure a place in the AI answer
In 57% of the cases where a brand's own page was on Google's first page for a question (216 of 378 across the five markets), none of the five AI engines named the brand. Brand-owned websites were 2 to 6% of the sources the answers cited. Retailer product pages were the largest identified source group in the Nordic markets in September, and Amazon appeared in 16% of UK AI answers.
Cases where a brand's own page was on Google's first page for a question, UK, Sweden, Norway, Finland and Denmark, September and October 2026. Source: Odyssey.
Matilda's view
This is the finding I would take to the SEO team, gently. A first-page Google position is still worth having, but in more than half the cases it bought no mention in the AI answer, and sometimes the answer named a competitor instead. Being the source and being the brand named in the answer are two different jobs, and the second belongs to the brand, PR, social and CX teams as much as to search.
Matilda Rydow, CCO, Odyssey
8. Sponsored and organic conversation give different pictures
In Sweden, sponsored posts were 22% of the monitored posts about beauty product brands but 53% of their estimated reach and 63% of the attributed PR value. In the UK they were 28% of posts and 20% of reach.
Sweden, monitored posts about beauty product brands on TikTok, YouTube, podcasts and broadcast, January to September 2026. Source: Odyssey and All Ears.
Matilda's view
I would be careful about reporting sponsored and organic posts as one number. Paid posts show what the brand bought, and organic posts show what people chose to say. In Sweden, sponsored posts were about a fifth of the posts and more than half of the reach, so a combined total can say more about the media budget than about the brand.
Matilda Rydow, CCO, Odyssey
9. Once paid posts are removed, organic conversation and AI answers point to the same brands
The two reach people at different moments: a post comes up as someone scrolls or listens, and an AI answer responds to a shopping question. Even so, brands with a larger share of organic conversation also tend to have a larger share of AI mentions, and in 19 of 20 combinations of market and category, AI mentions rank brands more like organic posts than like recognition of the brands' advertising. Averaged, the two give what the report calls share of earned: 19.9% for CeraVe in UK skin and body care, against 6.9% of buyers' preference, and 3.0% for Dove, against 17.0%.
Rank correlation of the brands' shares within each category's competitor set, five markets and four product categories. Conversation: organic posts, January to September 2026 (UK June to September). AI: five engines, September and October 2026. Source: Odyssey and All Ears.
Matilda's view
I like share of earned as a working number, on one condition: nobody puts it in a bonus scheme until we know it predicts something. The two halves are useful on their own. What the composite adds is a single line that can sit next to mind share, and the gap between the two for CeraVe and Dove is the kind of thing that starts a useful meeting.
Matilda Rydow, CCO, Odyssey
The study does not show whether AI mentions or first places lead to preference or sales. CeraVe leads the UK skin and body care answers, while its aided awareness in the survey fell from 55% to 42% between Q3 2025 and Q3 2026, and we cannot tell from the data whether the two are connected. Answering that needs repeated measurement of the same brands over time.
How the pieces fit together
Brand measurement is usually split across several studies, often bought from different suppliers in different years, and in large companies sometimes from a different supplier in each market:
Segmentation identifies the buyers the brand should win.
Positioning research describes what the brand should mean to them.
Communication evaluation tests whether campaigns land and are linked to the brand.
Brand tracking follows awareness, consideration, preference and associations over time.
Social listening and AI visibility show what is said about the brand and which brands AI answers recommend.
Media measurement, from attribution to marketing mix models, estimates what each channel adds to sales.
Each study can be done well and still use its own target group, competitors and customer needs, and then the results cannot be joined afterwards. We described four such gaps between brand and media measurement in Brand and media measurement still sit in silos. AI visibility and social listening add mismatches of their own: a question set that covers some needs far more than others, a tool that tracks brands picked from a keyword list, a listening report that adds paid and organic posts together.
Matilda's view
Most companies I talk to have more measurement than they can use. The gap is rarely a missing source.
Matilda Rydow, CCO, Odyssey
The fix starts with one description of the market: the buyers you want to win, what the brand should mean to them and the competitors it is up against. The survey supplies the demand spaces, their size and the competitor list, and every other source is read against them as shares of the same competitors.
Inside the tracker, the core measures stay the ones brand trackers are built on: unaided and aided awareness, consideration, preference, purchase and the associations buyers give each brand, asked the same way of the same population in every wave. How often to report depends on what the tracker feeds. A marketing mix model needs monthly results in the markets it covers, with samples large enough for a monthly figure, while quarterly brand reviews can use monthly fieldwork reported by quarter.
For CMOs and marketing leaders
Most marketing organisations buy a segmentation study, a positioning study, a tracker, social listening and an AI visibility tool. They usually arrive from different teams, each with its own competitor list and its own idea of the market, and then they cannot be read together. In a large company the split can run inside one discipline too, for example with a different tracking supplier in each market. The findings above show what that costs. The same brand can look strong in one source and weak in another, and a board that sees one number at a time has no way of telling which picture to trust. What is lost is the chain from segmentation to the position the brand should take, from a new position to evaluating whether it lands, and on to the early signals that show it moving. Each step only works when it uses the same buyers, needs and competitors as the step before.
The decisions that make the sources work together belong to you, and most of them are about the organisation: who owns brand measurement, whose budget pays for it, when the results are reviewed and which decisions they feed. A shared description of the market, covering the buyers you want to win, what the brand should mean to them and the competitors you are up against, is the base every supplier reports against. It only pays off when brand, insights, media and finance read the results together and act on them in the same forum. Until AI measures have been shown to relate to preference and sales, their place in 2027 is next to the survey measures for the same competitors, as context for the decisions you already make.
Questions to ask in 2027
Your team. Do all our brand numbers use the same competitors and the same customer needs?
Your team. Which decision would change if this number moved?
Your leadership team. Who owns brand measurement, who pays for it, and do brand, media and finance review the results in the same meeting?
Your AI visibility supplier. Which questions is the figure built on, and how many of them concern the need we want to own?
Every supplier. Can you report against our market definition, and what margin of error sits behind each figure?
Suppliers. Choose a brand survey that covers segmentation, positioning and tracking with the same respondents and competitors every month and in every market. If every new question needs a new data collection, the set-up is probably wrong. One continuous collection should let you run new analyses and drill into new cuts of the same data as the questions come up. Ask AI visibility suppliers to report by customer need, engine and market and to disclose their questions, and ask social listening suppliers to separate sponsored from organic posts in every market.
People. Give one person ownership of the market definition and the competitor list across all suppliers, teams and markets, so that everyone reads their results against the same view. Next to that person you need an analyst who can read survey margins of error and AI run-to-run variation side by side.
Do ask for the specification behind every number a budget decision rests on.
Do read AI visibility by customer need, next to unaided awareness, consideration and purchase.
Do put brand and media results on the same agenda, on the same schedule, with finance in the room.
Don't set an AI ranking as a board target before it has been tested against the survey.
Don't add a new data source before you have agreed which decision it will support.
Your first step for 2027: name one owner for brand measurement, agree one description of the market with every supplier, and set a quarterly review where brand, media and finance read the results together.
For insights teams
Your part of the system is the one the others depend on. Segmentation, positioning and the competitor list define the market that every other source should be read against, and AI answers and conversation become comparable with the survey only when they are expressed against the same competitors and the same customer needs. In 2027 the insights team holds the definitions that every other team reads its results against. Those definitions also carry the chain from segmentation to positioning, to evaluating whether a new position lands and to the early signals that show it moving.
The new sources also give you something back. AI answers and organic conversation show brands and needs that a fixed tracking list does not ask about, which makes them a practical way to review the list at set intervals. The time series stays intact while the list keeps up with the market. Before a new AI measure goes into regular reporting, it deserves the test you would give any new survey question: how it relates to the measures you already have, and what it adds.
Questions to ask in 2027
Your CMO. Which decisions should the tracker support in 2027, who reviews the results, and which competitors and needs do the decisions require?
Your AI visibility supplier. Can the question set be written from our demand spaces, with enough questions for each need?
Your social listening supplier. Can you report shares of the same competitor set we use in the survey?
Your own team. How does a new AI metric relate to unaided awareness, consideration and purchase over several quarters, and what does it add?
Suppliers. Prefer one continuous survey for segmentation, positioning and tracking, set up the same way in every market, over separate projects, so that the same respondents answer every part in every wave. A new question from the business should then lead to a new analysis of the data you already hold, with no new fieldwork. If the company uses a marketing mix model, the tracker needs to report monthly in the markets the model covers, with samples large enough for a monthly result. Ask AI visibility suppliers whether they can work from your question set and competitor list.
People. You need someone who can turn demand spaces into an AI question set, and an analyst who is at home with both survey statistics and AI run-to-run variation. An association measured among about 36 respondents in one quarter has a margin of error of about ±14 points, so someone has to decide how much data to pool before a comparison is made.
Do express every source as a share of the same competitor set.
Do review the competitor list at fixed intervals against the brands AI answers and conversation name most often.
Don't put an AI metric into regular reporting before you have tested it against the survey for several quarters.
Don't build a buyer-weighted AI total before each need has enough questions to be read on its own.
Your first step for 2027: map your segmentation, positioning study, campaign tests and tracker, and check whether they share one sample and one competitor list.
For brand and strategy teams
AI answers give brand teams a second reading of the position. Buyers and AI answers can link the same brand to different needs, and the difference is worth understanding. It can mean that what is published about the brand has drifted from what buyers associate it with, or that buyers have not yet followed a change the brand has made.
For a brand that is repositioning, the gap at the start is a baseline. It is one step in a chain that runs from segmentation to the position the brand should take, to evaluating whether the new position lands and to the early signals that show it moving, and the baseline can only be followed if every step uses the same buyers, needs and competitors. Whether AI answers move before buyers do is still an open question, so a change in the answers is best treated as a signal to test against later survey waves.
Leaders and challengers have different things to check. A category leader should look at its standing on the need it wants to own, since a strong overall rank can hide a weak one there. A challenger should check whether it is on the competitor lists that the category's trackers use at all.
Questions to ask in 2027
Your insights team. For which needs do AI answers name us, and how does that compare with the position buyers give us?
Your insights team. Which brands do AI answers name in our place on the need we want to own?
Your PR or content team. Which sources do the answers on our core need cite, and what do they say about us?
Your CMO. Which target position are we measuring progress towards in 2027?
Suppliers. Look for AI visibility measurement that shows the needs a brand is named for, the sources the answers cite and what they say, next to the survey's positioning map for the same competitors. The positioning map should come from the same continuous survey as the tracking, so that a new question about the position can be answered from data you already have.
People. Positioning work needs someone who reads AI answers as qualitative material: which claims, which sources and which competitors appear next to the brand. That skill sits closer to brand strategy and PR than to search.
Do define the target position, the questions that express it and the competitors that hold it today together with insights, media and PR, before you measure.
Do read a gap between AI answers and the survey as a question to investigate.
Don't count a change in AI answers as progress until the survey has moved as well.
Don't judge the brand on one overall AI rank.
Your first step for 2027: measure where AI answers place your brand on the need you want to own, and record it as your baseline for 2027.
For media and performance teams
Brand and media measurement are usually reviewed by different teams on different schedules, and neither report uses the other's figures. A mix model with sales as its only outcome then estimates short-term effects and holds no information about the brand. When awareness and consideration from the tracker enter the model, part of later sales can be credited to the demand that upper-funnel media created, and the recommended budget can change with it.
Campaign evaluation is also one step in a longer chain. It shows whether the position the brand has chosen lands with the buyers it wants to win, and that only works when the campaign test uses the same target group, needs and competitors as the segmentation and the tracker.
Social reporting needs the same separation. A total that adds sponsored and organic posts together can rise with influencer spend while unsolicited discussion stays flat, so the two belong on separate lines. A ratio between them describes the conversation, and it says something about return only once spend and outcome data are added.
Questions to ask in 2027
Your insights team. Can the tracker report monthly in the markets and target groups our mix model covers?
Your mix modelling partner. Can brand measures enter the model, and how often will the model and the brand measures be re-estimated together?
Your social listening supplier. Can you report sponsored and organic posts as separate lines in every market?
Your creative team. Which ads performed in the channels, and did the survey show that their message was understood?
Suppliers. If you use a marketing mix model, choose a partner that can take tracker data as an input and a tracker that can deliver it monthly. For influencer and social reporting, ask for sponsored and organic posts separately, with sentiment measured on the organic posts. Ideally campaign evaluation runs in the same continuous survey as the tracker, so that each new campaign is read against the same buyers and competitors without a separate study.
People. Let the people who buy media, produce ads for the channels and analyse the data plan against the same target and read the same reports. One person should be responsible for creative results in both the channels and the survey, so that a version with high completion and a weak message is caught.
Do read ad recognition, consideration and purchase next to media delivery for the same target group.
Do judge public response on the organic posts.
Don't call the ratio of organic to sponsored posts a return on investment.
Don't let media results be reviewed without the brand measures beside them.
Your first step for 2027: check whether your tracker can report monthly in the markets your mix model covers before you add brand measures to the model.
For AI search and SEO teams
Search work carries over to AI answers only in part. A strong page on Google does not mean the brand is named, and the answers lean on sources the brand does not own, such as retailers, editorial sites and forums. AI visibility therefore depends on far more than your own site. That also makes it a shared task for search, PR, content and brand, and someone has to decide who owns it and where the results are reviewed.
How much an AI visibility number tells you depends on the questions behind it. A ranking across many needs describes the question set as much as the brands, so the most useful thing you can give the rest of the company is a report by customer need, with the questions, engines and runs stated and mentions, first places and citations kept apart. Then the insights and brand teams can read AI visibility next to their own measures. Read by need, AI answers can also give an early signal of whether a new position is starting to show in what is published about the brand.
Questions to ask in 2027
Your AI visibility supplier. Which questions are asked, how many, for which needs, and how many times each?
Your AI visibility supplier. How do you collect each engine, now that Microsoft requires a signed-in account for Copilot and its terms bar automated access?
Your AI visibility supplier. Is a question tracked in five countries priced once or five times?
Your insights team. Can our AI report sit next to unaided awareness, consideration and positioning for the same competitors?
Suppliers. Require six things with every AI result: the question set and any weighting, results per engine, the market and language, the period and number of runs, whether the measure is mentions, first places or citations, and the competitors the shares are calculated on. ChatGPT Business can help with the analysis, since workspace admins can add custom MCP connections and query data from a tracker or AI visibility service in ChatGPT; Odyssey's MCP works with ChatGPT and Claude. It cannot measure how ChatGPT presents your brand to shoppers, because answers in your workspace draw on your account and company context.
People. Work with PR, product content and retailer teams on how the brand is described on the pages the answers cite, since being cited and being named are different outcomes. Someone in the team should read samples of the answers, because a mention can also be a recommendation against the brand.
Do report AI results by engine and customer need, with mentions, first places and citations as separate numbers.
Do report Google rankings and AI mentions side by side.
Don't present an AI share without its question set, engines and runs.
Don't measure your brand in AI answers with a handful of prompts from your own account.
Your first step for 2027: ask your supplier for the full question set behind your current AI visibility number, and map each question to a customer need.
For PR, communications and social teams
Organic conversation and AI answers turned out to describe a similar kind of public visibility. What people publish about a brand without being paid, and what AI answers say about it, rank brands in similar ways. That brings social listening closer to AI search than it has been, and it gives PR and communications a broad read on how visible a brand is in public. When the brand takes a new position, organic conversation and AI answers are among the first places to look for signs that it is landing, to be confirmed in later survey waves.
It also raises the bar for how conversation is reported. Sponsored and organic posts need separate lines, platform coverage needs to be stated before the figures are used in competitive claims, and sentiment says most about public response when it is measured on the organic posts.
Questions to ask in 2027
Your listening supplier. Which platforms do you cover in each market, and how do you classify sponsored content?
Your listening supplier. Can you tag posts with the same brand associations our survey uses?
Your AI search team. Which editorial sites, retailer pages and forums do AI answers cite for our core need?
Your insights team. Which brands lead organic conversation but are missing from our survey list?
Suppliers. Look for listening that covers spoken media as well as text, including TikTok, YouTube, podcasts and broadcast, and that separates sponsored from organic posts. Associations tagged in the survey's own terms let you compare what people publish with the position buyers give the brand. Keep the monitoring running on a fixed set of brands and platforms, so that a new question can be answered from what is already collected.
People. PR shapes two things at once, what people say about the brand and what AI answers cite. Someone in the team should follow which sources the answers draw on for your core needs and plan coverage with that in mind. The listening results then belong in the same review as the tracker and the AI report, read against the same competitors.
Do report organic and sponsored conversation as separate lines in every market.
Do state platform coverage before using conversation figures in competitive claims.
Don't add paid and organic mentions into one share of voice.
Don't count being cited in an AI answer as the same outcome as being named in it.
Your first step for 2027: ask your listening supplier to split every 2026 report into sponsored and organic posts, so that 2027 starts from a clean baseline.
Roles you may also need
Three more functions decide whether the measurement holds together.
Procurement. Buy against one brief that states the market definition, the competitor list and the customer needs, and score suppliers on whether they can report against it. Prefer agreements that give continuous data and room for new analyses over a new project for each new question. Ask every AI visibility supplier for the items in the supplier checklist before you compare prices.
Finance. If the company uses a marketing mix model, ask for brand measures in it, since a model with revenue as its only outcome cannot credit upper-funnel media with later sales. Agree with the CMO which brand measures count as leading indicators and which AI measures stay as context until they have been tested.
Data and analytics. Keep survey, conversation and AI data on the same brand and market codes so that they can be joined. With an API or MCP connection, more people will query the same data from BI tools, Claude or ChatGPT, which makes one shared competitor list even more important.
How to choose brand tracking, AI visibility and social listening suppliers
Whichever suppliers you choose, the same requirements make their results comparable.
The supplier checklist
Brand survey. One sample for segmentation, positioning and tracking, the same competitor list in every wave, the same set-up in every market, monthly reporting, a stated margin of error and room to run new analyses on the same data.
AI visibility. The question set and any weighting, results per engine, market and language, period and runs, the measure, and the competitors the shares are calculated on.
Social listening. Platform coverage per market, sponsored and organic posts reported separately, and shares of the same competitor set.
Marketing mix model. Brand measures as inputs or outcomes, re-estimated together with the tracker on a fixed schedule.
All of them. Reporting against your market definition and your competitor list.
A first quarter of 2027 for the whole company
January: agree the market. The CMO, brand and insights agree one description of the market and one competitor list, with a named owner.
February: brief the suppliers. Every measurement supplier receives the description and the checklist, and AI visibility moves to a question set built from your customer needs.
March: read the sources side by side. The first quarterly review, with brand, insights, media, PR and finance together, shows survey, conversation and AI results as shares of the same competitors, by customer need, with a list of gaps to investigate.
Common questions about brand tracking and AI visibility
How do you do brand tracking?
Start with the decisions the tracker should support, then define the market: the buyers you want to win, their needs and the competitors. Ask the same questions of the same population in every wave, with a competitor list that stays stable and is reviewed at fixed intervals. In 2027, read AI answers and conversation against the same competitors, as shares of the same set.
How do you measure AI search visibility?
Ask a fixed set of questions to each AI engine repeatedly, and count how often each brand is mentioned, named first and cited. Write the questions from your customers' needs, report results per engine and per need, and express them as shares of the same competitors your survey tracks. A handful of prompts from your own account is not a measurement.
What is a good AI visibility score?
There is no fixed benchmark, because the score depends on which questions were asked and which competitors are counted. Compare your share of AI mentions with your share of unaided awareness and consideration among the same competitors, need by need. In UK skin and body care, CeraVe had about 35% of the tracked brands' AI mentions and about 5% of their purchases.
How do you choose an AI visibility tool?
Ask how each number is made before you compare prices: the question set and the needs it covers, results per engine, the period and number of runs, and the competitors the shares are calculated on. Ask how each engine is collected, since Copilot now requires a signed-in account, and whether results can sit next to your survey data. The supplier checklist covers the rest.
What is the difference between brand tracking and social listening?
Brand tracking measures what a weighted sample of category buyers know, consider and prefer. Social listening counts what people publish, and in our data TikTok made up between 63% and 96% of the monitored conversation depending on the market. Read conversation as a share of the same competitors as the survey, with sponsored and organic posts separated.
One view of the market behind every number
Matilda's view
The brands that will read AI search well are the ones that already know which buyers they are after, what they want to stand for and whom they compete with, because those three decisions say which questions to ask the engines and which answers matter.
Matilda Rydow, CCO, Odyssey
Brand Signal puts survey data, spoken media and AI answers on one positioning map for the same competitors. See how we measure AI visibility by demand space, or bring the data into Claude or ChatGPT through the API and MCP. If you want to go through the questions in this guide with your own data, we are glad to do that with you.
Survey. Odyssey Beauty Signal in the UK, Sweden, Norway, Finland and Denmark. Online panel interviews with people aged 16 to 65 who had bought beauty products in the past six months, collected every month and weighted. 23,745 respondents in the 12 months to August 2026.
Conversation. Posts and episodes on TikTok, YouTube, podcasts and broadcast, monitored by All Ears from January to September 2026 in the Nordic markets and from June to September 2026 in the UK. Sponsored and organic posts are classified separately.
AI answers. 60 product questions and 60 retailer questions in each market, asked of ChatGPT, Gemini, Copilot, Google AI Mode and Google AI Overviews in September and October 2026. Google search results for the same questions were recorded in the same runs.
Comparisons. Each figure is a brand's share of a defined set of competitors in its category and market. The full method and its limitations are described in the report.
Matilda has spent her career in marketing technology and media measurement, and in recent years on how they connect to the brand, in the technology and in the organisation.