AI keyword research for D2C brands.

What is different about AI keyword research

Classical keyword research built lists of two-to-four-word queries with clean volume and difficulty scores. The search interface was a text box and the result was ten blue links. AI keyword research works against a different surface: conversational input, generative answers, and selective citations. Queries are longer. Qualifiers stack. Follow-up turns shift intent inside a single conversation. Volume data is opaque. The research job changes from "which queries have volume and rank potential" to "which queries are being answered in AI, which sources get cited, and where does the catalog have a claim."

The overlap with classical SEO is large but not total. Intent research, topic clustering, and content planning still benefit from Semrush, Ahrefs, and Google Search Console. The new layer is citation research: for each priority query, which engines answer it, which sources are cited, and whether the brand's PDPs or blog posts are among them. See the What is GEO guide for the broader framework.

How buyers actually phrase shopping queries to AI

Observable patterns from public AI chat interfaces as of Q1 2026 show several consistent shapes. Buyers phrase queries as full sentences with context: "I'm doing a long weekend of backcountry skiing in late February, I run hot, what's the best merino base layer under $120?" They include budgets naturally. They mention use cases, body types, seasons, and constraints. They ask follow-up questions that reference earlier turns. They often ask the engine to compare two or three options rather than name the best one outright.

This is closer to a conversation with a knowledgeable salesperson than a keyword search. D2C brands that write PDPs and buying guides with that shape in mind produce content AI engines can lift directly into answers. See the optimize content for AI search guide.

Stacked qualifiers

Stacked-qualifier queries are the hallmark of AI shopping prompts. "Best merino base layer for backcountry skiing under $120" packs product type, material, use case, and price into a single string. Buyers type these queries naturally to AI engines because the engines handle them. On Google, they would type "merino base layer" and filter manually.

For a D2C catalog, stacked qualifiers are opportunity. The long-tail shape means there is less competition, and the specific product that matches the full qualifier stack tends to win the citation. A brand whose PDPs explicitly cover price, material, and use case signals in plain-language passages has a structural advantage on these queries. See the stacked qualifiers glossary entry.

Intent drift in AI conversations

A single AI conversation often shifts intent. A buyer might start with "what's the best merino base layer under $120?", move to "what thickness should I get for late winter?", then to "which of those ships in time for my trip next weekend?" That is three distinct intents in one conversation: discovery, specification, and logistics. Each turn can surface a different set of cited sources.

For keyword research, this means a flat keyword list undersells the opportunity. Mapping typical conversation arcs — discovery, comparison, specification, logistics, purchase, post-purchase — gives a content team a way to plan content that can be cited at more than one turn in the conversation.

Keyword lists optimize for the first thing a buyer types. Conversation arcs optimize for the third thing they ask, which is often where a citation decides the purchase.

Brand-less queries: the neutral-intent majority

A large share of AI shopping queries contain no brand name at all. "Best merino base layer for backcountry skiing under $120" is neutral-intent. The engine decides which brands to surface. Observable behavior across ChatGPT, Perplexity, and Google AI Mode as of Q1 2026 suggests that neutral-intent queries return three to seven cited sources, often blending retailer PDPs, independent reviews, and editorial buying guides.

This is the zone where D2C brands have the most to gain or lose. Ranking well on brand-name queries ("merino base layer brand X review") is defensive. Winning neutral-intent queries is offensive — it grows share of answer against competitors the buyer has not named yet. A good research plan prioritizes neutral-intent queries in every category the brand wants to win.

Prompt research workflows

Practical workflows combine three inputs. First, logs of actual customer prompts when available (CX transcripts, support tickets, on-site search queries as proxies). Second, AI engines themselves — asking ChatGPT and Perplexity "what are the ten most common questions a buyer asks before purchasing a merino base layer" produces usable starter lists when treated skeptically. Third, classical keyword tools as an intent cross-check.

The output is a set of per-category prompt libraries, each with twenty to fifty conversational queries tagged by intent (discovery, comparison, specification, logistics). This library feeds both content planning and the eCommerce Insights tracking configuration, which can monitor prompts for citation presence over time.

Tools for AI keyword research, honest assessment

Available tooling as of Q1 2026. Traditional SEO tools — Semrush, Ahrefs, SE Ranking, Google Search Console — remain useful for volume, intent, and classical-SERP context. AI-visibility platforms — eCommerce Insights, Profound, Brandlight — vary in whether they support custom prompt libraries; eCommerce Insights does. AI engines themselves, used with structured prompts, serve as a lightweight research input. No tool today produces clean AI-specific volume data comparable to Google Keyword Planner. Anyone selling that number should be questioned. See Google's guidance at developers.google.com/search.

Per-engine keyword strategy

Engines behave differently on the same query. Perplexity Shopping leans heavily on source citations and often surfaces two to seven PDPs directly. ChatGPT with shopping tools tends to synthesize more and cite less aggressively unless the user asks for sources. Google AI Mode and AI Overviews often pull from Google's shopping feed and from the classical web index. Gemini integrates tightly with Google's shopping data. Claude operates more conservatively, with cited sources emerging more often on explicit "show sources" prompts. These are as-of-Q1-2026 observations and will shift.

The implication for research: a prompt library should be run against at least ChatGPT, Perplexity, and Google AI Mode at minimum. Engine-specific lists that prioritize where the brand has the most to gain tend to produce tighter quarterly roadmaps.

Seasonal and intent-context drift

Query volume and intent drift seasonally in ways that classical keyword tools capture poorly for AI surfaces. Gift categories spike November and December. Outdoor categories split by season and region. Home categories shift around tax season and spring. AI engines respond to current context, so queries about "gift ideas for a sibling who just started rock climbing" behave differently in October and in May.

Quarterly research refreshes catch drift. Build a standing agenda item: at each quarter turn, re-run the top twenty prompts per priority category and note which sources are now being cited. Track the competitor PDPs that appear for the first time.

Mapping AI queries to PDPs vs collection pages

Mapping research output to the catalog involves a decision per query: does this prompt land best on a specific PDP, a collection page, a buying guide, or the blog? Highly specific stacked-qualifier queries often land on a specific SKU's PDP. Broader category queries may land on a collection or buying guide. Comparison queries benefit from a dedicated compare or buying guide. The mapping shapes content work for the quarter.

Common mistakes

Three mistakes recur. First, chasing volume instead of citability — high-volume head terms often get answered from parametric memory with few cited sources, so the optimization ceiling is low. Second, treating AI queries like Google queries and running optimization off two-word head terms. Third, researching in one engine only and assuming the results generalize. Each engine reads the same query differently; the research plan has to reflect that. See the eCommerce Insights glossary and schema guide for the measurement vocabulary.

Key takeaways

• AI queries are conversational, long, and stacked with qualifiers.
• Neutral-intent queries are where D2C brands have the most to gain.
• Plan for intent drift across turns, not just single-query optimization.
• Prioritize citability and intent over raw search volume.
• Research per engine; one engine's answer does not generalize.

Related guides