Best ChatGPT Prompts for eCommerce SEO

eCommerce SEO is not a writing problem. It is a scale problem. You do not have ten pages that need better copy — you have four thousand product URLs, six hundred of them near-duplicates of each other, a faceted navigation generating another eighty thousand crawlable combinations, and a manufacturer feed that gave every competitor in your category the identical description you are using.

That is exactly the situation AI is good at. It is also exactly the situation Google spent 2026 penalising. The March 2026 core update named scaled content abuse as a primary target, and the sites that got hit were the ones that generated thousands of pages and published them unread.

So these 30 prompts are built around a specific discipline: AI produces the draft, a human clears the gate, and the gate is a real one. Prompt 15 exists solely to be that gate. Skip it and the rest of this article becomes a liability rather than a system.

The Rule That Governs All of This

Google does not penalise content for being AI-generated. It penalises unoriginal, low-value content produced at scale to manipulate rankings, regardless of how it was made. The distinction is not the tool — it is whether the page has a reason to exist.

What that means practically:

• AI-drafted, human-reviewed product copy is fine. A first draft that gets fact-checked against real specs and edited for voice is normal publishing.
• Raw, unedited AI output published across hundreds of SKUs is not. That is the exact pattern enforcement targets.
• Legitimate scale is allowed. Catalogues, programmatic pages and local landers survive when they carry real, accurate, differentiated data.
• The differentiator is information gain. If your page contains nothing the manufacturer feed does not already say, it has no reason to rank.

Eight Rules That Make These Prompts Work

1. Feed real data or expect invented data. Every product prompt below takes a spec block. A model given no specs will confidently invent dimensions, materials and compatibility — and a wrong spec in a product description is a returns problem, not an SEO problem.
2. Category pages carry the head terms. The most common structural mistake in eCommerce SEO is optimising product pages for terms that a category page should own. Get the page-type mapping right before you write anything.
3. Never let it invent a search volume. Models hallucinate keyword metrics fluently. Use them to generate and classify candidates; get the numbers from a real tool.
4. One page, one intent. Cannibalisation in eCommerce usually comes from a category and a subcategory chasing the same query, not from two blog posts.
5. Templates need variables that actually vary. A title template that produces near-identical strings across 500 SKUs is duplicate content with extra steps.
6. Technical issues outrank copy issues. If faceted navigation is burning your crawl budget, better product descriptions will not save you. Fix the crawl first.
7. Write for the person mid-purchase. Someone comparing two products wants specifics, trade-offs and what is in the box — not adjectives.
8. Nothing publishes unreviewed. Prompt 15 is the gate. Use it.

Keyword and Intent Research (Prompts 1–5)

1. The Category Keyword Map

Maps queries to page types before anything gets written.

Act as an ecommerce SEO strategist. Map search demand to page types for my catalogue.

STORE: [WHAT YOU SELL, AND TO WHOM]
CATALOGUE STRUCTURE: [YOUR CURRENT CATEGORY TREE, PASTED]
PRODUCT COUNT: [ROUGHLY HOW MANY SKUS]
PRIMARY MARKETS: [COUNTRIES / LANGUAGES]

For the category tree above, produce:
1. The head term each top-level category should own, and why that term belongs to a category page rather than a product page
2. For each subcategory: the mid-tail cluster it should target
3. Which terms belong on a product page instead, and the test for telling them apart
4. Any term in my tree that two pages would compete for, flagged as a cannibalisation risk
5. Gaps — demand my current tree has no page for, and whether that needs a new category, a filter landing page, or a guide

Rules:
- Do NOT estimate search volumes or difficulty scores. Generate and classify candidates only; I will pull metrics from a real tool
- Classify every term by intent: transactional, commercial investigation, or informational
- Where a term is ambiguous, say so rather than guessing

Output as a table: term, intent, target page type, target URL, notes.

2. The Intent Classifier

Intent decides page type. Page type decides everything else.

Classify these keywords by search intent and tell me what page each one needs.

KEYWORDS: [PASTE YOUR LIST]
MY CATALOGUE: [BRIEF DESCRIPTION OF WHAT YOU SELL]

For each keyword, return:
1. Intent: transactional, commercial investigation, informational, or navigational
2. The page type that should target it: category, subcategory, filtered collection, product, buying guide, comparison, or none
3. What a searcher using this term expects to see in the first screen
4. Whether my catalogue can actually satisfy it, or whether ranking would produce a bounce

Then flag separately:
- Terms where the intent is mixed and a single page cannot serve both readings
- Terms that look transactional but are really research queries
- Terms I should not target at all because the click would not convert

Be strict about the difference between commercial investigation and transactional. Most ecommerce sites lose traffic by pointing a product page at a query that wanted a comparison.

3. The Modifier Expansion

Expand this head term into the long-tail modifier space my catalogue can serve.

HEAD TERM: [THE TERM]
PRODUCT ATTRIBUTES AVAILABLE: [LIST THE REAL ATTRIBUTES IN YOUR DATA — size, colour, material, compatibility, capacity, price band, use case, brand]
CATALOGUE DEPTH: [HOW MANY PRODUCTS SIT UNDER THIS TERM]

Generate long-tail variations grouped by modifier type:
- Attribute modifiers (from the real attributes above only)
- Use-case modifiers
- Audience modifiers
- Comparison and alternative modifiers
- Problem or symptom modifiers
- Purchase-stage modifiers (best, cheap, near me, reviews, vs)

For each group, tell me:
1. Which of these deserve their own indexable page and which should stay as filters
2. The minimum product count that justifies a dedicated page
3. Which would produce a thin page my catalogue cannot fill

Rules: only generate modifiers my stated attributes can actually support. Do not invent product attributes I did not list — a landing page for an attribute I do not stock is a dead end.

4. The Competitor Gap Analysis

Analyse where competitors capture demand that my site does not.

MY CATEGORY STRUCTURE: [PASTE]
COMPETITOR STRUCTURES: [PASTE THEIR CATEGORY TREES / NAVIGATION AS YOU OBSERVE THEM]
MY PRODUCT RANGE: [WHAT YOU STOCK]

Identify:
1. Page types they have that I do not — categories, guides, comparison pages, filtered landers
2. Where their taxonomy is more granular than mine, and whether that granularity is justified by demand or is just clutter
3. Content formats they rank with that I have no equivalent of
4. Where my range is stronger but my structure hides it

Then give me:
- The five gaps most worth closing, ranked by effort against likely return
- Which gaps are traps — pages that exist for SEO reasons and probably do not convert
- What I could do better rather than merely matching

Base this only on the structures I pasted. Do not assert what any competitor ranks for or how much traffic they get — you cannot verify that, and I will check it in a real tool.

5. The Seasonal Demand Plan

Build a seasonal SEO calendar for my catalogue.

CATEGORIES: [YOUR MAIN CATEGORIES]
MARKET: [COUNTRY / REGION]
KNOWN SEASONAL PATTERNS: [ANYTHING YOU ALREADY OBSERVE IN YOUR OWN SALES DATA]
LEAD TIME: [HOW LONG YOUR CONTENT AND DEV CYCLES TAKE]

Produce:
1. A month-by-month map of which categories peak when, based on the patterns I described plus general retail seasonality
2. For each peak: how many weeks before it the page needs to be live and indexed
3. Which pages should be permanent and re-used annually versus genuinely temporary
4. The URL strategy for recurring seasonal pages — keep one evergreen URL and update it, rather than creating a new dated URL each year, and explain why
5. What to do with a seasonal page out of season

Flag clearly which parts come from my own stated data and which are general assumptions I should validate against my analytics before committing budget.

Category and Collection Pages (Prompts 6–9)

6. The Category Page Brief

Write the brief for a category page that should rank for a head term.

CATEGORY: [NAME]
TARGET TERM: [THE HEAD TERM]
PRODUCTS IN CATEGORY: [HOW MANY, AND THE RANGE — price band, key attributes, brands]
COMPETING PAGES: [WHAT CURRENTLY RANKS, AS YOU OBSERVE IT]
CUSTOMER QUESTIONS: [WHAT BUYERS ACTUALLY ASK, FROM SUPPORT OR REVIEWS]

Produce a brief covering:
1. The single job this page does for a searcher
2. What belongs above the fold — products, or orienting content, and why for this specific term
3. The filters and sorts that should be visible immediately, based on how buyers in this category actually decide
4. What supporting copy is genuinely useful, where it sits, and what it must contain to add information the product grid does not
5. The internal links this page needs, in both directions
6. The FAQ block, drawn from the real customer questions above

Rules:
- Do not recommend a wall of keyword-stuffed text below the product grid. If supporting copy would not help a buyer, say it should not exist
- Specify a word count only where the content justifies it
- Name what makes this page different from the subcategory pages beneath it

7. The Category Copy

Write the supporting copy for this category page.

CATEGORY: [NAME]
TARGET TERM: [PRIMARY] and [SECONDARY TERMS]
RANGE: [WHAT IS ACTUALLY IN THIS CATEGORY — brands, price spread, key differences between options]
BUYER DECISION: [WHAT A CUSTOMER IS ACTUALLY CHOOSING BETWEEN]
BRAND VOICE: [PASTE TWO PARAGRAPHS OF YOUR EXISTING COPY]

Produce:
1. An H1 and a one-sentence intro that orients rather than sells
2. 120-180 words of above-grid copy that helps someone choose, not just describes the category
3. 200-300 words of below-grid copy covering the decision criteria that matter in this category
4. A meta title under 60 characters and meta description under 155
5. Three FAQ pairs answering real purchase questions

Rules:
- Every claim must come from the range data I gave you. Invent nothing about materials, standards, compatibility or performance
- No filler openings. Do not begin with a sentence that restates the category name back to the reader
- The below-grid copy must contain at least one genuinely useful decision criterion a competitor page probably omits
- Match my voice sample

Then tell me which sentence in your draft is the weakest and why.

8. The Cannibalisation Check

Check whether these pages compete with each other.

PAGES: [FOR EACH — URL, page type, H1, target term, and a short summary of its content]
GSC DATA IF AVAILABLE: [QUERIES AND IMPRESSIONS PER PAGE]

Assess:
1. Which pages target overlapping intent, and how severe the overlap is
2. For each conflict: which page should own the term, with the reasoning
3. Whether the losing page should be consolidated, redirected, re-targeted at a different term, or left alone
4. Where the overlap is actually fine because the intents differ more than the wording suggests

Then recommend, per conflict:
- The specific action, and the risk of taking it
- What to change on the page that keeps the term
- What to monitor for four weeks after the change, and what result would mean reverse it

Be conservative with redirect recommendations. Consolidating two pages that were serving different intents loses more traffic than it saves, so tell me when to leave things alone.

9. The Filter Page Decision

Which facet combinations deserve indexable pages — the decision that quietly determines your crawl health.

Decide which of my faceted filter combinations should become indexable landing pages.

FACETS AVAILABLE: [LIST EVERY FILTER — brand, colour, size, price, material, feature, rating]
CATALOGUE SIZE: [SKU COUNT]
EXAMPLE COMBINATIONS: [LIST SOME REAL ONES]
KNOWN SEARCH DEMAND: [ANY TERMS YOU KNOW PEOPLE SEARCH]

For each facet and combination type, decide:
1. INDEX — real standalone search demand, enough products to fill the page, and a distinct intent
2. NOINDEX, FOLLOW — useful to users, no search demand, links should still pass
3. BLOCK FROM CRAWL — no user or search value, and it multiplies the URL space

Give me:
- The rule set, expressed so a developer can implement it
- The minimum product count a filtered page needs before it earns indexation
- Which combinations to never allow indexed, with the reasoning
- How many URLs each rule roughly removes from the crawlable space
- The canonical strategy for the noindex tier

Rules: default to restrictive. Combinatorial facets are the single most common cause of index bloat in ecommerce, and every additional indexable combination has to justify the crawl cost it creates.

Product Pages (Prompts 10–15)

10. The Product Description

Write a product description from the real specification below.

PRODUCT: [NAME]
SPECIFICATIONS: [PASTE THE COMPLETE REAL SPEC SHEET — dimensions, materials, weight, compatibility, contents, warranty, certifications]
PRICE AND POSITION: [WHERE IT SITS IN YOUR RANGE]
WHO BUYS IT: [THE ACTUAL CUSTOMER AND WHAT THEY NEED IT FOR]
COMMON QUESTIONS: [FROM SUPPORT TICKETS OR REVIEWS]
BRAND VOICE: [PASTE A SAMPLE]

Produce:
1. A 40-60 word opening that states what it is and who it suits
2. A benefits section that maps each benefit to the specific spec that delivers it
3. A specification table using only the data above
4. A "what is in the box" list
5. Three FAQ pairs from the real questions
6. Meta title under 60 characters, meta description under 155

Absolute rules:
- Use ONLY the specifications provided. If something would strengthen the copy but is not in the spec, list it under MISSING instead of inventing it
- No unverifiable performance claims, no comparative claims about competitors, no compliance or safety claims not present in the source data
- Do not use superlatives the spec cannot support

End with the MISSING list so I know what to get from the supplier.

11. The Manufacturer Feed Rewrite

The single highest-value fix in most catalogues — your competitors are all running the same text.

Rewrite this manufacturer description so it is original and more useful than the version every competitor is running.

MANUFACTURER COPY: [PASTE IT]
WHAT I KNOW BEYOND THE FEED: [YOUR OWN PHOTOGRAPHY NOTES, REVIEW THEMES, RETURN REASONS, SUPPORT QUESTIONS, HOW IT COMPARES TO OTHERS YOU STOCK]
CUSTOMER: [WHO BUYS IT FROM YOU SPECIFICALLY]

Produce a rewrite that:
1. Preserves every factual specification exactly
2. Reorganises around the buying decision rather than the manufacturer marketing angle
3. Adds the information gain that only I have — fit notes, what people actually use it for, what surprises buyers, who it is not right for
4. Removes manufacturer boilerplate that says nothing

Then tell me explicitly:
- Which parts of the rewrite are genuinely new information versus rephrased feed content
- Whether there is enough information gain here to justify the page ranking, or whether I need to add photography, video or review content before this page can compete

Be honest on that last point. A rephrased feed description is still a near-duplicate in substance, and rewording alone does not create a reason to rank.

12. The Title and Meta Template

Design title and meta description templates for my product pages at scale.

CATALOGUE: [WHAT YOU SELL]
AVAILABLE DATA FIELDS: [LIST EVERY FIELD IN YOUR PRODUCT DATA — brand, model, key attribute, size, colour, capacity, price, rating, stock status]
SKU COUNT: [HOW MANY PAGES THIS APPLIES TO]
CATEGORY EXAMPLES: [THREE REAL PRODUCTS WITH THEIR FIELD VALUES]

Produce:
1. A title template using only fields that genuinely vary between products, under 60 characters when populated
2. A meta description template under 155 characters when populated
3. Rendered examples using the three real products above
4. Fallback rules for products with missing field values
5. Category-specific variants where one template will not serve the whole catalogue

Then stress-test your own template:
- Show me what it produces for three near-identical SKUs that differ only in colour
- If the outputs are near-duplicates, redesign it so they are not
- Tell me which products in a catalogue like mine would need hand-written titles instead

A template that produces 500 nearly identical strings is a duplicate-content problem, not a scaling win.

13. The Product FAQ

Turn real customer questions into a product FAQ block.

PRODUCT: [NAME AND SPEC]
REAL QUESTIONS: [PASTE FROM SUPPORT TICKETS, REVIEWS, LIVE CHAT, OR PRE-SALES EMAILS]
RETURN REASONS: [WHY THIS PRODUCT COMES BACK]

Produce 6-8 FAQ pairs that:
1. Answer the questions actually asked, in the words customers used
2. Prioritise questions that precede a purchase decision over post-purchase ones
3. Address the return reasons directly and honestly — this reduces returns even though it may reduce conversions
4. Stay factual to the spec

For each answer:
- Lead with the direct answer in the first sentence
- Keep to 40-70 words
- No marketing language in an answer to a practical question

Then tell me which of these questions suggests the product page itself is unclear, because an FAQ that exists to patch confusing copy is a symptom rather than a fix.

14. The Variant Strategy

Recommend how to handle product variants for search.

PRODUCT FAMILY: [DESCRIBE]
VARIANT AXES: [WHAT VARIES — colour, size, capacity, material]
VARIANT COUNT: [HOW MANY COMBINATIONS]
SEARCH BEHAVIOUR: [DO PEOPLE SEARCH FOR SPECIFIC VARIANTS — e.g. by colour — as far as you know]
CURRENT SETUP: [HOW YOUR PLATFORM HANDLES IT NOW]

Decide, with reasoning:
1. One page for the family with variant selectors, or separate indexable pages per variant
2. Where the split point is if the answer is mixed — e.g. separate pages for capacity, selectors for colour
3. Canonical strategy for whichever model you recommend
4. URL structure
5. How structured data should represent the variants
6. What happens to a variant page when that variant is discontinued

Then give me the case against your recommendation, and the specific signal in my analytics that would tell me the other model was right.

15. The Scaled Content QA Gate

Run this on every batch before publishing. This is the prompt that keeps the other twenty-nine out of trouble.

Act as a quality reviewer. Audit this batch of AI-drafted product content before it publishes.

BATCH: [PASTE 5-10 GENERATED DESCRIPTIONS, WITH THE SOURCE SPEC FOR EACH]
CATALOGUE CONTEXT: [WHAT ELSE IS ON THE SITE]

For every item, check:
1. FACTUAL — does every stated spec match the source data exactly? List any claim not supported by the spec
2. INVENTED — flag any dimension, material, certification, compatibility or performance claim that appears in the draft but not in the source
3. DUPLICATE — how similar is each draft to the others in this batch? Identify sentences that repeat across items with only the product name changed
4. INFORMATION GAIN — does each page contain anything a manufacturer feed does not already say? If not, say so plainly
5. VOICE — does it read as written by a person who knows the product, or as template output?
6. CLAIMS — any comparative, health, safety, environmental or compliance claim that needs substantiation

Then give me:
- A PUBLISH / EDIT / REJECT verdict per item, with the reason
- The systemic problem in this batch, if there is one — a repeated failure across items means the generation prompt needs fixing, not the outputs
- An estimate of what proportion of this batch is genuinely differentiated

Be harsh. Google's scaled content abuse policy targets exactly this pattern — large volumes of unoriginal, low-value pages regardless of how they were produced. If this batch would fail that test, say so directly rather than softening it.

Technical SEO (Prompts 16–20)

16. The Product Schema Audit

Audit and specify structured data for my product pages.

CURRENT MARKUP: [PASTE YOUR EXISTING JSON-LD, OR STATE WHAT YOU HAVE]
AVAILABLE DATA: [WHICH FIELDS YOU CAN POPULATE — price, currency, availability, condition, SKU, GTIN, brand, reviews, ratings, shipping, returns]
PLATFORM: [SHOPIFY / WOOCOMMERCE / MAGENTO / CUSTOM]

Produce:
1. Which schema types this page should carry and why — Product, Offer, AggregateRating, Review, BreadcrumbList, FAQPage
2. Required versus recommended properties for each, and which of them I can actually populate today
3. Fields I am missing that would unlock richer results
4. Where my current markup is wrong, incomplete, or contradicts what is visible on the page
5. The correct handling for variants, price ranges and out-of-stock items

Rules:
- Structured data must match what a user sees on the page. Flag anything marked up but not visible
- Never mark up ratings that do not exist on the page or that came from anywhere other than genuine customer reviews
- Note where a property is optional but materially improves eligibility for rich results

Give the output as a field-by-field specification a developer can implement, plus the validation steps to run afterwards.

17. The Faceted Navigation Crawl Plan

Build a crawl control plan for my faceted navigation.

FACETS: [EVERY FILTER AND ITS VALUE COUNT]
URL PATTERN: [HOW FILTERED URLS ARE CONSTRUCTED — parameters or paths]
CATALOGUE SIZE: [SKU COUNT]
CURRENT CONTROLS: [WHAT ROBOTS.TXT, CANONICALS, NOINDEX OR PARAMETER HANDLING YOU HAVE NOW]
CRAWL SYMPTOMS: [WHAT GSC CRAWL STATS AND INDEX COVERAGE SHOW]

Produce:
1. A rough calculation of how many URLs my current facet setup can generate
2. Which controls to apply at which tier — robots.txt disallow, noindex, canonical, nofollow on links, or parameter handling
3. The exact rules, written so a developer can implement them without interpretation
4. The order to roll them out, safest first
5. What to monitor after each change, and the signal that means roll it back

Be specific about the difference between blocking a crawl and removing an index entry, and about why blocking a URL in robots.txt does not remove it from the index if it is already there. Getting that order wrong is the most common way sites make this worse while trying to fix it.

18. The Out-of-Stock Policy

Design my policy for out-of-stock and discontinued products.

CATALOGUE: [WHAT YOU SELL]
TYPICAL PATTERNS: [HOW OFTEN THINGS GO OUT OF STOCK, FOR HOW LONG, AND WHETHER THEY RETURN]
CURRENT HANDLING: [WHAT YOU DO NOW]
TRAFFIC PROFILE: [DO DISCONTINUED PRODUCT PAGES STILL GET SEARCH TRAFFIC]

Give me a decision framework covering:
1. Temporarily out of stock — what the page does, what schema says, what the user sees
2. Discontinued with a direct replacement
3. Discontinued with no replacement but ongoing search demand
4. Discontinued with no demand
5. Seasonal items that return annually

For each, specify: HTTP status, indexation, canonical, structured data availability value, internal linking, and what the page should show the customer.

Rules:
- Never 404 a page that still earns traffic and has a relevant destination
- Never redirect to a category page as a blanket rule — an irrelevant redirect is treated as a soft 404 and frustrates users
- Explain the trade-off between keeping a page for its links and the user experience of landing on something unbuyable

19. The Pagination and Canonical Review

Review my pagination and canonicalisation setup.

STRUCTURE: [HOW CATEGORY PAGINATION WORKS — numbered pages, load more, infinite scroll]
URL PATTERN: [HOW PAGE 2+ URLS LOOK]
CURRENT CANONICALS: [WHAT THEY POINT TO]
VIEW-ALL: [WHETHER ONE EXISTS]
PRODUCTS PER PAGE: [NUMBER]

Assess:
1. Whether my current canonical treatment of paginated pages is correct or is hiding products from indexation
2. Whether infinite scroll or load-more is preventing crawler access to products beyond page one, and how to check
3. Whether a view-all page is appropriate for my catalogue size
4. How paginated pages should handle titles, metas and self-referencing canonicals
5. Whether products deep in pagination are effectively orphaned

Then give me:
- The specific changes, in priority order
- How to verify crawler access to page-two products before and after
- The risk of each change

Flag any place where my setup is likely causing products to be excluded from the index entirely, since that is a revenue problem rather than a ranking one.

20. The Ecommerce Core Web Vitals Triage

Triage my page speed problems by revenue impact.

TEMPLATES AND THEIR METRICS: [FOR EACH PAGE TYPE — LCP, INP, CLS, and mobile versus desktop]
TRAFFIC AND CONVERSION BY TEMPLATE: [WHICH TEMPLATES CARRY THE REVENUE]
PLATFORM AND CONSTRAINTS: [WHAT YOU CAN AND CANNOT CHANGE]
KNOWN CAUSES: [THIRD-PARTY SCRIPTS, IMAGE HANDLING, APPS INSTALLED]

Produce:
1. Which template to fix first, weighted by traffic and conversion rather than by how bad the score is
2. The likely causes of each failing metric for that specific template type
3. Fixes ranked by impact against implementation effort
4. Which third-party scripts are most likely responsible, and how to confirm before removing anything
5. What is realistically achievable on my platform versus what would need a rebuild

Be clear about which of these changes affect real user experience and which only move the lab score. Also state plainly where speed is not my actual ranking problem — if the diagnosis is that content or crawl issues matter more here, say so rather than optimising the wrong thing.

Content and Topical Authority (Prompts 21–24)

21. The Buying Guide Brief

Brief a buying guide that ranks and sends people to products.

TOPIC: [THE GUIDE SUBJECT]
TARGET TERM: [PRIMARY TERM AND INTENT]
MY RANGE: [WHAT YOU STOCK IN THIS CATEGORY, WITH THE REAL DIFFERENCES BETWEEN OPTIONS]
BUYER CONFUSION: [WHAT PEOPLE GET WRONG WHEN CHOOSING]
EXPERTISE I HAVE: [WHAT YOUR TEAM ACTUALLY KNOWS — support experience, testing, returns data]

Produce a brief covering:
1. The single question the guide answers
2. Section structure, with what each section must contain
3. Where product recommendations belong and how many is too many
4. The decision framework to give the reader — the criteria that actually matter, in priority order
5. What to include that a manufacturer or a generic affiliate site could not write, drawn from my stated expertise
6. Internal links out to categories and products, and the anchor text for each
7. Schema markup appropriate to the format

Rules:
- The guide must be genuinely useful to someone who buys nothing
- Recommend against including a product where my range genuinely is not the right answer, and say what to do instead
- No fabricated testing claims. If I have not tested these, the guide must not imply that I have

22. The Comparison Page

Structure a comparison page between products or categories I stock.

COMPARING: [A versus B, or the set]
REAL DIFFERENCES: [PASTE THE ACTUAL SPEC DIFFERENCES]
WHO EACH SUITS: [YOUR HONEST VIEW]
MARGIN OR STOCK CONSIDERATIONS: [IF ANY — state them so I can see whether they are biasing the output]

Produce:
1. A comparison table of the attributes that genuinely affect the decision, not every spec
2. A verdict per use case: for X, choose A; for Y, choose B
3. Where the cheaper option is genuinely sufficient, stated plainly
4. The trade-off that is easy to miss
5. Title, meta and H1 targeting the comparison query
6. Internal links to both product pages

Rules:
- Do not manufacture a difference where two products are effectively equivalent — say they are equivalent and give a different deciding factor
- Do not let the margin note above influence the recommendation. If it did, say so
- A comparison page that always concludes "buy the expensive one" reads as sales copy and will not earn links or trust

23. The Topical Cluster Map

Map the content cluster that supports one of my commercial categories.

CATEGORY: [THE MONEY PAGE THIS SUPPORTS]
TARGET TERM: [WHAT THE CATEGORY WANTS TO RANK FOR]
EXISTING CONTENT: [WHAT YOU ALREADY HAVE, WITH URLS]
EXPERTISE AND ASSETS: [WHAT YOU CAN CREDIBLY PRODUCE]

Produce:
1. The pillar page and what it must cover
2. Supporting articles, each with its target term, intent and the specific question it answers
3. The internal linking structure between them and back to the commercial page
4. Which existing content fits the cluster, which should be updated, and which should be consolidated or removed
5. Priority order by likely commercial return, not by search volume

Rules:
- Every piece must have a reason to exist beyond covering a keyword
- Flag any topic where I would have nothing original to say — those are the pieces that will not perform and will dilute the cluster
- Keep the cluster to a size I can actually maintain; an unmaintained cluster ages badly and drags the pillar with it

24. The AI Search Visibility Pass

Optimise this page to be usable as a source by AI search and assistants.

PAGE: [PASTE THE CONTENT, OR THE URL AND A SUMMARY]
PAGE TYPE: [CATEGORY / PRODUCT / GUIDE]
QUERIES IT SHOULD SERVE: [LIST THEM]

Assess and improve:
1. Whether the page answers its core question in a single extractable passage near the top
2. Whether facts are stated in self-contained sentences that survive being quoted without surrounding context
3. Whether specifications, prices and availability are in structured, machine-readable form as well as prose
4. Whether comparative and superlative claims are attributable and specific rather than vague
5. Whether the page states clearly who produced it and on what basis

Then produce:
- The rewritten opening passage
- A list of facts on the page that are currently only implied and should be stated explicitly
- The structured data that would make the key facts unambiguous

Rules: this is about clarity and extractability, not about writing for machines at the expense of people. If a change would make the page worse for a human reader, do not recommend it. Do not add claims to make the page more quotable — an extractable falsehood is worse than an unquoted truth.

Internal Linking and Architecture (Prompts 25–27)

25. The Internal Link Plan

Design the internal linking for my category and product pages.

SITE STRUCTURE: [YOUR TREE]
PRIORITY PAGES: [THE COMMERCIAL PAGES THAT MATTER MOST]
EXISTING LINKING: [HOW PAGES LINK NOW — nav, breadcrumbs, related products, footer, body copy]
CONTENT ASSETS: [GUIDES AND ARTICLES YOU HAVE]

Produce:
1. Where each priority page should receive links from, in priority order
2. The linking pattern between category, subcategory and product levels
3. How related-product modules should be populated — by attribute, by co-purchase, by price band — and which serves both users and crawlers best here
4. Which links belong in navigation versus in body content, and why the distinction matters
5. Footer links: what earns a place and what is dilution

Then flag:
- Pages that are over-linked relative to their commercial value
- Priority pages that are structurally too deep, with the click depth from the homepage
- Where the navigation is passing weight to pages that do not need it

26. The Anchor Text Distribution

Review my internal anchor text.

LINKS: [PASTE — source page, destination page, current anchor text]
DESTINATION TARGETS: [WHAT EACH DESTINATION WANTS TO RANK FOR]

Assess:
1. Whether anchors describe the destination accurately
2. Where the same anchor points at different destinations, or different anchors at the same one inconsistently
3. Over-optimisation — the same exact-match anchor repeated at scale
4. Uninformative anchors that waste the signal ("click here", "this page", "read more")
5. Accessibility: whether each anchor makes sense read out of context

Produce a revised anchor for every link that needs one, plus:
- The distribution I should aim for across exact, partial, branded and natural-language anchors
- Which links to leave exactly as they are
- Where varying the anchor would help and where consistency matters more

Prioritise clarity for a human reader over keyword placement. An anchor that reads awkwardly to a person is a poor link regardless of its keyword.

27. The Orphan Page Recovery

Find and fix pages that have no internal links pointing to them.

CRAWL DATA: [PASTE YOUR ORPHANED OR LOW-INLINK URLS]
SITEMAP COUNT VS INDEXED COUNT: [THE NUMBERS]
CATEGORY STRUCTURE: [YOUR TREE]

For each orphaned page, decide:
1. Should it exist at all — is there demand, and does it serve a purpose
2. If yes: where in the structure does it belong, and which pages should link to it
3. If no: consolidate, redirect, or remove, with the specific target
4. Why it became orphaned — the pattern usually points at a structural cause, such as products only reachable through deep pagination or discontinued category branches

Then tell me:
- The systemic cause behind the largest group of orphans
- The structural fix that prevents recurrence, rather than the one-off link additions
- Which orphans are actually fine to leave unlinked

A one-off cleanup that does not address why pages orphaned will need doing again in six months.

Measurement and Recovery (Prompts 28–30)

28. The Search Console Query Analysis

Analyse my Search Console data and tell me where the opportunity is.

DATA: [PASTE QUERY-LEVEL EXPORT — query, page, impressions, clicks, CTR, average position]
PERIOD: [DATE RANGE]
CATEGORY CONTEXT: [WHAT THESE PAGES SELL]

Identify:
1. High-impression, low-CTR queries where the title and meta are the constraint rather than the ranking
2. Queries ranking in positions 5-15 where a modest improvement would matter commercially
3. Queries where the ranking page is the wrong page for the intent
4. Queries with commercial intent that my catalogue can serve but that rank poorly
5. Queries I rank for that I should not chase, because they will not convert

For the top opportunities, give me:
- The specific change to make
- The expected mechanism — better CTR, better relevance, or a different page entirely
- How long to wait before judging it

Rules: work only from the data I pasted. Do not estimate volumes or difficulty. Where the data is insufficient to draw a conclusion, say so rather than inferring one. Prioritise by commercial value, not by impression count.

29. The Cannibalisation Diagnosis

Diagnose why my rankings for this term are unstable.

TERM: [THE QUERY]
PAGES RANKING FOR IT OVER TIME: [PASTE THE URLS AND DATES IF THE RANKING PAGE HAS SWAPPED]
GSC DATA: [IMPRESSIONS AND POSITIONS PER PAGE FOR THIS QUERY]
PAGE DETAILS: [FOR EACH CANDIDATE — title, H1, content summary, internal links in]

Determine:
1. Whether this is genuine cannibalisation or normal ranking fluctuation
2. Which page Google appears to prefer, and what signal is driving that
3. Where my own signals are contradictory — titles, internal anchors, and content all pointing different ways
4. The specific fix: consolidate, differentiate, re-target, or adjust internal linking

Then:
- State the confidence level of your diagnosis and what additional data would raise it
- Give the least-destructive fix first
- Say what to measure over the next month and what would indicate the fix worked

Do not recommend a merge or redirect unless the evidence genuinely supports it. Ranking swaps between two pages are often normal, and an unnecessary consolidation loses one page's traffic permanently.

30. The Traffic Drop Post-Mortem

Diagnose a drop in organic traffic.

THE DROP: [WHEN IT STARTED, HOW STEEP, WHICH METRIC — clicks, impressions, or both]
SCOPE: [SITEWIDE, OR SPECIFIC TEMPLATES, CATEGORIES OR PAGES]
GSC DATA: [IMPRESSIONS, CLICKS, POSITION BEFORE AND AFTER]
INDEX COVERAGE CHANGES: [ANY]
WHAT CHANGED ON MY SIDE: [DEPLOYS, MIGRATIONS, CONTENT PUBLISHED, PLATFORM OR APP CHANGES, ROBOTS OR CANONICAL EDITS]
KNOWN ALGORITHM ACTIVITY: [ANY UPDATES IN THE WINDOW]

Work through, in this order:
1. Is this a tracking or reporting artefact rather than a real drop
2. Is it a technical availability problem — pages deindexed, blocked, returning errors, or slow to the point of failure
3. Is it a self-inflicted change — the timeline against my own deploys
4. Is it demand-side — seasonality or a genuine fall in search interest for these terms
5. Is it a SERP change — more ads, AI answers, or new features reducing clicks at unchanged rankings
6. Only then: is it an algorithmic quality assessment

For each, state what evidence would confirm or rule it out, and what my data already indicates.

Then give me the single most likely cause, your confidence in it, the diagnostic to run next, and the recovery path if the diagnosis holds. If impressions held steady while clicks fell, say clearly that this points at the SERP rather than at my rankings.

Running These as a System

1. Map before you write — prompts 1 and 2 decide which page type owns which term. Everything downstream depends on getting this right.
2. Fix the crawl — prompts 9, 17 and 19. Better copy on pages Google cannot reach efficiently is wasted work.
3. Build the category layer — prompts 6 and 7. These carry the head terms.
4. Then products at scale — prompts 10 to 14, with prompt 15 on every batch, without exception.
5. Add the content layer — prompts 21 to 24 for the queries no product page can serve.
6. Wire it together — prompts 25 to 27.
7. Measure and correct — prompts 28 to 30 on a monthly cycle.

Mistakes That Cost Rankings

1. Publishing generated descriptions unreviewed. The failure mode Google explicitly targets. Prompt 15 exists for this and it is the one step nobody wants to do.
2. Letting the model invent specifications. A wrong dimension is a return, a bad review and a refund — the SEO cost is the least of it.
3. Optimising product pages for category terms. The most common structural error in eCommerce SEO.
4. Indexing faceted combinations by default. Index bloat builds quietly and is painful to unwind.
5. Trusting AI keyword metrics. Models generate plausible search volumes that are entirely fabricated. Generate candidates with AI, get numbers from a tool.
6. Rewording the manufacturer feed and calling it original. Rephrasing is not information gain.
7. 404ing discontinued products that still earn traffic and links.
8. Blanket-redirecting dead products to the homepage or a category. Treated as soft 404s and it annoys customers.
9. Marking up ratings that are not on the page. A structured data penalty is avoidable and self-inflicted.
10. Fixing content when the problem is technical. Diagnose before you write.

Where to Run These

The drafting and analysis prompts work in any capable model, though the diagnostic ones — 28, 29 and 30 — reward a frontier reasoning model that will show its working and admit when your data is insufficient. For anything needing live SERP or competitor data, an agentic tool that can genuinely browse beats a model reasoning from training data, and our entrepreneur prompt set covers the same discipline applied to financial analysis.

For the rest of the eCommerce stack: OpenArt product video prompts cover the visual side of product pages, ChatGPT prompts for AI ads cover paid creative, and how to use ChatGPT for content creation covers the editorial workflow these plug into.

Keep Reading

More prompt sets and workflows: ChatGPT prompts for AI ads, how to use ChatGPT for content creation, OpenArt product video prompts for ecommerce, 23+ AI prompts for entrepreneurs, what ChatGPT Work is, and luxury brand creatives with ChatGPT. Or browse all guides and prompts on PromptsRush.