MMarketing Against The Grain
← All frameworks
Marketing

AI Answer-Engine Share-of-Voice Playbook

Map buyer questions, publish niche answers, and earn model-specific citations

Difficulty
Moderate
Time to result
~months to results
Steps
6
Confidence
97%

The playbook replaces keyword-only SEO planning with question-based visibility management. First, model the ideal customer and generate the many conversational ways that customer might ask about a product, problem, or category. Test those questions across target assistants to establish the brand's current share of voice. Then create narrow, answer-ready pages and FAQs addressing the missing variations. Improve external corroboration by earning mentions on sources that each assistant frequently uses, recognizing that ChatGPT and Gemini may rely on different licensing partners and websites. Re-run the question set to measure whether answer inclusion improves. Because referral clicks remain small even when visibility rises sharply, the primary output is trusted presence inside AI-generated answers rather than conventional organic traffic.

Origin

Extracted from Marketing Against the Grain, where the hosts described the AEO program used to increase HubSpot's AI-assistant share of voice by roughly 800%.

Core principles

  • 01Follow changing buyer behavior rather than defending old channels
  • 02Measure visibility across realistic buyer questions
  • 03Give language models precise content they can incorporate into answers
  • 04Match citation efforts to each model's preferred sources
  • 05Treat AI-search visibility more like brand building than direct response

How to run it

  1. 1

    Model the ideal customer

    Use customer evidence and AI to create a detailed representation of the buyer, including their language, needs, and buying context.

    Pro tip Ground the profile in real customer conversations rather than demographic assumptions alone.

    Watch out A generic customer profile will produce generic questions and weak content priorities.

  2. 2

    Generate conversational questions

    Ask AI to enumerate the different ways the customer would discuss products, services, use cases, and comparisons. Preserve natural phrasing instead of reducing everything to keywords.

    Pro tip Generate approximately 100 meaningful variations when the category supports that depth.

    Watch out Do not assume the three to five keywords used in traditional SEO represent conversational search.

  3. 3

    Establish share of voice

    Run the question set through relevant AI assistants and record how often the brand appears in their answers. Use this as the baseline for later comparisons.

    Pro tip Break results down by assistant and buyer-intent category.

    Watch out Referral traffic alone will substantially understate AI visibility.

  4. 4

    Fill answer gaps

    Create narrow pages, FAQs, examples, and product explanations that directly answer underrepresented questions. Include distinctive evidence such as original data or customer cases where possible.

    Pro tip Write self-contained passages that an assistant can accurately incorporate into an answer.

    Watch out Mass-producing undifferentiated AI copy adds noise without giving models a reason to cite the brand.

  5. 5

    Build model-specific citations

    Identify the sites each target assistant commonly uses, then earn credible mentions on those sources. A citation is a relevant mention, not merely a traditional backlink.

    Pro tip Investigate licensing relationships, such as the relationship between ChatGPT and Reddit.

    Watch out A citation strategy optimized for one assistant may not transfer to another.

  6. 6

    Measure and iterate

    Re-run the same questions after content and citation changes, compare share of voice, and prioritize the remaining gaps.

    Pro tip Track visibility and referral traffic separately.

    Watch out Do not abandon the program merely because attributable clicks remain small.

In the wild

HubSpot's AEO program

HubSpot built a dedicated AI-engine-optimization pod, measured how often it appeared in answers about its market, and used niche content, FAQs, and citations to improve coverage. The company reported much greater visibility and referral growth, while acknowledging that AI referral traffic was still tiny compared with historic organic traffic.

Approximately 800% higher AI-assistant share of voice and 1,400% higher AI referral traffic were reported.

Assistant-specific community coverage

A B2B software company finds that ChatGPT rarely mentions it in comparison questions. Its team identifies Reddit as an influential source, participates transparently in relevant discussions, publishes supporting comparison pages, and then re-tests the original question set.

The company gains more frequent inclusion in ChatGPT answers without treating Reddit participation as backlink spam.

Common mistakes

Tracking search volume instead of lost clicks

Conversational interfaces can increase the number of searches while reducing visits to websites. Visibility and click disappearance must therefore be measured separately.

Publishing one broad product page

One page optimized around a few keywords cannot answer the many conversational variations buyers use with AI assistants.

Using one citation plan for every model

Different assistants rely on different partnerships and sources, so a universal outreach list can miss the sources that actually influence each model.

Is it for you?

Best for

It is best for B2B companies whose educational or transactional search traffic is being cannibalized by AI answers.

Not ideal for

It is not ideal for teams expecting immediate, precisely attributable performance-marketing returns.

From the transcript

you can use AI to build an ICP ideal customer profile

Kieran Flanagan · 18:00

Now you need like a hundred variations of that product page because of the way people ask questions about that product or service.

Kieran Flanagan · 19:00

you really want to understand what LLM you're optimizing for and then who their license agreements are with

Kieran Flanagan · 20:00

From the episode

AI Just Broke Marketing (And What You Need to Do Now)