Skip to content
Industries / Beauty

How discovery has changed in beauty

Shoppers no longer move from ten blue links to a decision on their own. Search, AI answers, comparison, recommendation and action now happen across several answer surfaces, and your brand is either inside that answer or it is not.

60%of Google searches beginning with a question word returned an AI summary, so an answer is drawn before a link is chosen.Pew Research Center, 68,879 searches by 900 US adults, March 2025
58%lower clickthrough rate for the top-ranking page once an AI Overview sits above it.Ahrefs, 300,000 keywords via Search Console, 2026
86-90%of the pages an AI Overview cites already rank in the top 10, so search and answers share evidence.Ahrefs, 2025

Search has not gone away. Its job has widened into the evidence system the answer layer draws on.

The shopper journey

What does a shopper actually do before deciding?

No named persona, one real purchase decision. Each stage is a different question, and each question is now answered somewhere you may not be measuring.

  1. 1NeedI want something that suits my skin and lasts.
  2. 2DiscoverWhat kinds of product solve this?
  3. 3ResearchWhich shade and formula match me?
  4. 4CompareIs the expensive one actually better?
  5. 5DecideWill this work on my skin type?
  6. 6ActBuy it, from somewhere genuine.
Answer surfaces

One question, now asked in six places

The same beauty question now appears across search, answers, assistants and agents, and each surface reaches it through different evidence.

Which foundation should I buy for my skin?
ThenGoogle SearchTen blue links, ranked.
NowGoogle AI OverviewAn answer above the links.
NowChatGPTRetrieval plus model context.
NowGeminiThe Google ecosystem, synthesised.
NowPerplexityCitation-led web synthesis.
NowAgents and assistantsThe answer becomes an action.

One question. Many surfaces. Different evidence.

Query fan-out

The shopper does not want ten links

One question hides six smaller ones. An answer that resolves them wins the decision, and an answer that cannot resolve them from your pages resolves them from someone else’s.

Which foundation should I buy for my skin?
Shade and undertone match
Skin type and finish
Wear time and transfer
Ingredients and sensitivities
Cruelty-free and vegan status
Where to buy it genuine
Tracked prompts

Prompts Ansyra starts you on in this category

These are seeded when you set up a Beauty brand, then run against every engine on your plan. You add your own from there.

best long wear matte foundation for oily skin
how to match foundation shade to your undertone at home
drugstore mascara vs luxury mascara which is worth the price
which foundation shade suits a warm undertone
where to buy authentic serums and sunscreen
SEO, still

Search has not stopped mattering. Its job changed.

The answer layer sits above the links, but it is largely built from what already ranks. Search now decides which pages are even eligible to be drawn from, which makes it the entry condition rather than the finish line.

DiscoverableYour pages appear when people search the category at all.Ranking for the head terms and the long tail.
RelevantThe page matches the intent behind the query, not just its words.Guides, comparisons, calculators, plan detail.
TrustworthyAuthority, transparency and proof an engine can weigh.Ratings, disclosures, verifiable numbers.
Click-worthyThe title and snippet earn the click that is still available.Clear value, no bait, credible framing.
What the open web can still prove
RankWhere you sit for a query, and whether that is moving.
ImpressionsHow much demand your pages are actually in front of.
Click-throughWhether the snippet earns the visit once shown.
SessionsWhat arrives, which is the part an answer layer takes from you first.
The five disciplines

Five questions, five outcomes

Each term maps to a different moment in the journey. Together they are one loop, not five separate programmes.

The five disciplines of AI discovery, and what each one optimises
DisciplineThe question it answersWhere it plays outOutcome
SEOCan shoppers find us?Traditional search resultsFound
AEODo we appear in the answer?AI overviews and answer boxesAnswered
GEOAre we trusted across engines?Generative enginesTrusted
AIODo we stay useful in conversation?Multi-turn conversationsPreferred
AXOCan an agent act with us?Assistants and agentsActionable
SEOCan shoppers find us?Rank in traditional search results.OutcomeFound
AEODo we appear in the answer?Be cited or recommended in AI overviews.OutcomeAnswered
GEOAre we trusted across engines?Be visible across ChatGPT, Gemini, Perplexity and more.OutcomeTrusted
AIODo we stay useful in conversation?Deliver answers that hold up over follow-up questions.OutcomePreferred
AXOCan an agent act with us?Let assistants and agents complete the task.OutcomeActionable
The answer ladder

Mentioned is not the same as recommended

A yes or no hides the difference between being named in passing and being put forward as the choice. Ansyra grades the position, so you can see which rung you are on and what moves you up.

  1. 1AbsentThe answer does not contain you.The answer recommends products and yours is not one.
  2. 2MentionedNamed, but not linked or explained.Your brand is named with no product or shade attached.
  3. 3CitedNamed with a source the reader can open.Your shade guide is cited as the source for a match.
  4. 4ComparedPresent in the shortlist shoppers weigh.Your product is in the comparison the shopper is weighing.
  5. 5RecommendedPut forward as a leading choice.Your product is recommended for this skin type and undertone.
What moves you up a rung
Answer-ready pagesContent that answers the question directly instead of circling it.
Factual evidenceData and proof points an engine can cite without hedging.
Source authorityCitations earned from the places your category already trusts.
Comparison coveragePresence in the side-by-sides where the choice is actually made.
Answer-ready pages

What does a strong beauty page need?

The same structure serves the shopper, the crawler and the answer engine. These are the blocks an engine looks for when it decides whether it can answer from you or has to go elsewhere.

1Shade dataUndertone and depth per shade, so a match can be computed rather than eyeballed.
2The full ingredient listINCI in text, which is what a sensitivity question is answered from.
3Skin-type fitWhich skin types and concerns the formula is built for.
4Wear and finishLongevity, transfer and finish claims, with whatever supports them.
5Where to buy it genuineAuthorised stockists, so the answer does not send people to a counterfeit.
Why an engine can use it
Clear structureLogical sections in a predictable order, so a machine can find the part that answers.
Factual evidenceNumbers with their source and date attached, which is what survives a verification pass.
Schema-ready dataKey details marked up so they can be parsed rather than inferred from prose.
Direct answersQuestion-shaped blocks mapped to the questions people actually ask.
And the four blocks that make one answer-ready
1Direct answer40 to 60 wordsA plain-language answer that goes straight to the point.Instant clarity
2Comparison tableSide by sideThe two or three axes the decision actually turns on.Easy evaluation
3Evidence blockSource and dateThe proof points behind the claim, each checkable.Earned trust
4Follow-up questionsWhat comes nextThe question the reader asks after this one.Continued discovery
The evidence behind the answer

AI has to verify what is said about you

For beauty, engines reconcile your own pages against independent sources before naming you. The strength of that evidence set is what decides whether you are cited or hedged away.

Product and shade pagesShades, finishes, ingredients and claims, structured for matching.
Retailer and marketplace listingsWhere the product is actually sold, priced and rated.
Ingredient referencesINCI and ingredient databases engines use to answer sensitivity questions.
Beauty publishersEditorial roundups and tested-by pieces that answers cite heavily.
Creator reviews and videoSwatches and wear tests, which carry unusual weight in this category.
Community forumsWhere real skin-type experience and dupes are debated.
Trusted is five things, not one
Breadth of sourcesHow many independent places say it.
FreshnessWhether the newest version is the one being read.
ConsistencyWhether those sources agree with each other and with you.
AuthorityHow much weight the publishers carry in your category.
ClarityWhether the facts are stated so they can be lifted without interpretation.
No regulator pack

Ansyra ships no regulator rule pack for beauty today. Content still runs through the general honesty and evidence checks that apply to every brand, and you can attach your own rules. We say this plainly rather than implying screening that does not happen.

Engine variance

Why do different engines give different answers?

One beauty question can resolve differently on each engine, because each one reaches for a different set of sources first. Cross-engine visibility matters more than winning any single engine.

Google AI OverviewTypical evidenceSearch index plus the open webHeavily overlaps what already ranks.
ChatGPTTypical evidenceRetrieval plus model contextLeans on well-structured reference pages.
GeminiTypical evidenceThe Google ecosystem plus the webPulls in video and knowledge surfaces.
PerplexityTypical evidenceCitation-led synthesisShows its sources, so citations matter most.
Failure modes

Where does the beauty journey break?

Each of these is a moment the shopper was ready and the answer could not carry them. They are fixable, and they are the work Ansyra ranks for you.

Breakpoint 01Shades an engine cannot matchIf shade names carry no undertone or depth data, the assistant cannot answer the one question that matters.
Breakpoint 02Counterfeit listings outrank youGrey-market sellers become the cited source, and the answer sends shoppers to a fake.
Breakpoint 03No stock or store pathThe shopper is sold and the answer cannot say where to buy it today.
Breakpoint 04A discontinued shade recommendedOld pages stay crawlable, so the assistant confidently names something you no longer make.
What fixing them is aiming at
Fewer dead endsKeep the conversation moving toward the task.
Stronger contextPreserve what the reader already told the assistant.
Safer recommendationsProtect trust, and stay inside the rules.
Agent readiness

Agent-readiness needs four layers

The end state is not a click. Without these four, an assistant cannot finish the task safely, and the shopper is handed back to a form.

01Readable truthCurrent products, terms, limits and exclusions.Accurate, structured, current.
02Callable toolsSearch, quote, book, support.Functions an agent can trigger and trust.
03Explicit policiesPermissions, limits and compliance rules.Guardrails that keep actions safe.
04Recovery pathsHuman support, fallback and confirmation.A clear way to escalate and confirm.
The taskMatch a shade and add it to a basket
  1. 1Capture skin type, undertone and finish
  2. 2Check the range for a match
  3. 3Compare formulas on wear and ingredients
  4. 4Check price and availability
  5. 5Confirm the basket or store
Who owns this

Who owns AI visibility?

Being found, answered, trusted, preferred and actionable are five different jobs sitting in five different teams. They only add up if they are measuring the same thing.

SEOCrawlability and structure
  • Indexability
  • Site architecture
  • Discoverability
ContentAnswer-ready pages
  • Topical depth
  • Clarity and completeness
  • Intent alignment
PRAuthority and coverage
  • Brand mentions
  • Publisher quality
  • Source diversity
ProductTools and data
  • Data quality
  • Structured signals
  • Product experience
AnalyticsProof and movement
  • Visibility trends
  • Conversion lift
  • Impact measurement
One shared measurement layer underneath
AI visibilityAcross engines and models
CitationsWho cites what, and where
Share of answerAcross topics and rivals
SentimentTrust and perception
ActionsTraction and outcomes
The platform

Ansyra shows where the journey breaks

Prompts, answer visibility, sources, competitors and actions, connected. Five questions, one view, and a ranked list of what to fix next, in any of 29 markets and the buyer’s own language.

01What are shoppers asking?The real prompts in your category, ranked by demand and intent.
02Are we present in the answer?Visibility, position and sentiment across the engines your plan measures.
03Which sources shape the answer?The domains and citations the engines actually drew on.
04Where do competitors win?Rival presence and mention share on the same prompts.
05What should we fix next?A ranked worklist, with the evidence behind each item.
Getting started

What do the first 90 days look like?

Baseline the reality before promising automation. Awareness only becomes an outcome when the loop repeats.

01Weeks 1 to 3
Baseline the current reality

Establish how you show up today across answers and engines, before changing anything.

  • Baseline prompts
  • Rankings across engines
  • Citations and sources
  • Competitor mentions
02Weeks 4 to 7
Fix the gaps that block answers

Make the pages answer-ready, with the facts, sources and structure an engine can verify.

  • Answer-ready pages
  • Complete facts and stats
  • Source and citation gaps
  • Schema and structure
03Weeks 8 to 12
Measure the impact and scale

Re-measure the same prompts, then widen to the next set of journeys and topics.

  • Measure visibility lift
  • Review answer share
  • Expand to priority journeys
  • Document and repeat
Questions shoppers ask

Beauty and AI answers, in short

How does an AI assistant decide which foundation to recommend for oily skin?

It assembles an answer from whatever it can read about the formula, not from ad spend. Shade and undertone data, skin-type fit, wear claims, ingredient lists and retailer listings all feed it. Where that detail is missing or locked inside an image, a rival's readable page gets matched instead.

What does a beauty shopper weigh before buying a foundation?

Shade match comes first, then skin type and finish, wear time, ingredients and sensitivities, cruelty-free status, and where to buy it genuine. Any one of those left unanswered is where the shopper stops. Assistants work through the same list, using whichever source can answer each part.

What should a beauty product page include so an assistant can match a shade?

Undertone and depth for every shade, in text, so a match can be computed rather than eyeballed. Add the full INCI list, the skin types and concerns the formula is built for, wear and finish claims with what supports them, and authorised stockists so the answer does not point at a counterfeit.

How do you tell whether a beauty brand is being recommended by AI assistants?

Run the prompts shoppers actually type, repeatedly, and record what comes back. Track whether the brand is absent, named without a product, cited as the source for a shade match, present in the comparison, or recommended for a stated skin type. Answers vary between runs, so single checks prove little.

See how AI answers beauty questions about you

Start with a free trial, or take a walkthrough on your own prompts and the shoppers you sell to.