Skip to content
SEOAIExpert

ChatGPT SEO

Getting recommended inside a conversation, not ranked on a page

In an assistant conversation there is no results page to compete on. There is one response, a handful of named options and a short list of citations. ChatGPT SEO is the work of becoming one of those named options.

Book a Strategy CallExplore Our Approach
Competitive field
3–5 named brands
Decision moment
Before any click
Lever
Clarity + corroboration
Assistant responseIllustrative
  • ChatGPT
  • Search-enabled
  • Follow-up asked
Which running shoe brands are best for wide feet and flat arches?

Generated answer

Three brands come up consistently for wide fits with arch support. Each publishes explicit width and support specifications, which makes them easy to compare.

  • Brand A88/100

    Publishes width and arch specs per model

  • Brand B61/100

    Strong reviews, thin technical detail

  • Brand C34/100

    Fit guidance buried in a PDF

Sources citedbrand-a.com/fit-guidebrand-a.com/wide-fitreview-site.com

Assistant visibility score

74/100

Conversational behaviour

One prompt replaces an entire research session

Assistant users do not run five searches. They describe a situation, receive a shortlist, then interrogate it — narrowing on price, fit, compatibility or availability inside the same thread.

  • Prompts contain constraints, not keywords.
  • Follow-ups reuse the sources already retrieved.
  • A brand missing from the first response rarely enters later.
  • Specific, checkable detail wins over persuasive language.
Prompt to shortlistSequenced
  1. 01

    Situation described

    “Wide feet, flat arches, under £150, run three times a week.”

  2. 02

    Constraints extracted

    Width, support type, price ceiling, usage frequency.

  3. 03

    Retrieval triggered

    Sources fetched for the specific constraint set.

  4. 04

    Shortlist formed

    Brands with checkable specifications survive.

  5. 05

    Follow-up handled

    The same sources are reused to narrow the choice.

Source selection

Why a brand gets named — and why one gets skipped

Selection is closer to fact-checking than ranking. The response needs statements it can stand behind, attributed to a source that will not contradict itself.

  • Gets named

    Easy to verify

    • Explicit specifications in text
    • Consistent naming across the site
    • Clear who a product is and is not for
    • Stable facts confirmed off-site
    • Availability and pricing in structured data
  • Gets skipped

    Expensive to verify

    • Claims without measurable detail
    • Specifications only inside images or PDFs
    • Different product names in different places
    • Content requiring interaction to appear
    • Marketing copy with no stated constraint

Access and retrievability

Assistants can only quote what they are allowed to fetch

Before content quality matters, access must be intact. This is the most common silent blocker we find on ecommerce sites.

Assistant access diagnosticIllustrative
  • Crawler directives permit AI user agentsReview
  • Key content present in server HTMLPartial
  • Specifications available as textGaps
  • Product schema completeStrong
  • Response time under loadGood
  • Canonical consistencyStrong

Entity clarity

The assistant has to know what your brand is before it can suggest it

We define the brand as an entity with attributes a model can restate: what you sell, who you serve, where you ship, what makes a product suitable, and how your naming works.

Entity definition for assistantsIllustrative

Product range

Named, specified

Use cases

Who it suits

Fit & sizing

Measured

Entity

Your brand

Markets served

Shipping scope

Comparisons

Stated verdicts

External sources

Consistent facts

Illustrative entity map. Each node is a fact set an assistant can restate confidently in a recommendation.

Content work

Six content patterns that earn assistant citations

  • 01

    Constraint-first buying guidance

    Guidance organised by the constraints people actually state — budget, size, compatibility, use frequency — with a direct recommendation per constraint.

    Prompt matchHigh
  • 02

    Specification blocks in text

    Every measurable attribute written as text on the page and mirrored in Product schema, never left inside an image or spec-sheet download.

    ExtractabilityHigh
  • 03

    Explicit unsuitability

    Stating who a product is not for is unusually quotable, because assistants need exclusion criteria to narrow a shortlist.

    Citation valueStrong
  • 04

    Named comparisons

    Comparison content that reaches a verdict per use case rather than presenting a neutral feature grid.

    Answer fitStrong
  • 05

    Question-shaped headings

    One question per heading, answered in the first two sentences beneath it, so a single passage can stand alone.

    Passage qualityHigh
  • 06

    Stable fact sheets

    A canonical place for shipping, returns, warranty and availability facts, matching what third-party sources say.

    ConsistencyHigh

Prompt coverage

Coverage is planned against prompts, not keyword volume

We build a prompt set from the way customers describe problems, then map each prompt to the page and passage responsible for answering it.

Prompt coverage map
Illustrative coverage map used to prioritise content work.
Prompt patternIntentResponsible pageCoverage
“Best X for [constraint]”ShortlistCategory buying guideCovered
“Is X good for [situation]?”ValidationProduct page FAQCovered
“X vs Y for [use case]”ComparisonComparison pagePartial
“What size X do I need?”SpecificationSizing guidanceGap
“Where can I buy X in [market]?”AvailabilityShipping factsCovered
“Cheaper alternative to X”SubstitutionCategory alternativesGap

Related work

Where ChatGPT SEO sits inside AI search

Conversational assistants are one surface. The underlying clarity work compounds across every answer engine.

  • AI SEOthe broader discipline this page sits inside.
  • AEOoptimising for direct question answers.
  • AEO Auditdiagnose answer readiness first.
  1. 01

    Access

    Fetchable, rendered, fast

  2. 02

    Clarity

    Entities and specs explicit

  3. 03

    Answers

    Quotable passages

  4. 04

    Corroboration

    Facts agree off-site

  5. 05

    Recommendation

    Named in responses

Measurement

How assistant visibility is reported

Illustrative demo data showing the reporting model. Assistant answers vary between sessions, so we report trends across a fixed prompt set rather than single observations.

Illustrative demo data
  • Prompt appearance rate41%Share of monitored prompts naming the brand
  • Owned-page citations26%Citations pointing to your own domain
  • Shortlist position1.8Average position among named brands
  • AI referral sessions+34%Sessions from assistant referrers

Figures shown are illustrative interface data, not client results.

FAQ

ChatGPT SEO questions

You cannot control the output, but you can change the inputs it retrieves and how easily they can be verified. When specifications, use cases and constraints are stated explicitly on accessible pages and corroborated elsewhere, a brand becomes far more likely to be named and cited than a competitor whose detail is unavailable or inconsistent.

Yes. Responses vary with phrasing, retrieval and model updates, which is why single observations are not evidence. We monitor a fixed prompt set repeatedly and report appearance rate and citation share as trends.

If your directives block the user agents that fetch pages for search-enabled responses, your content cannot be retrieved or cited. Blocking is a legitimate business decision, but it should be a deliberate one — we audit directives first and make the trade-off explicit.

It shares the foundation and diverges in emphasis. Both need crawlable, rendered, well-structured pages. Assistant visibility additionally rewards self-contained passages, measurable specifications and stated verdicts, because those are what a response can safely quote.

Category buying guidance, product specification content, comparison pages with clear verdicts, and stable fact pages covering sizing, shipping and returns. These carry the constraints that conversational prompts are built from.

Through prompt-level tracking — appearance rate, citation share and average shortlist position — read alongside AI referral sessions and assisted conversions, so visibility gains connect to commercial outcomes.

Prompt-level review

Find out what assistants say about your category today

We build a prompt set for your products, record what current responses recommend, and show the specific access, clarity and content gaps keeping your brand out of the shortlist.

Book a Strategy Call

No obligation. A senior strategist reviews your store before the call.