Build SEO comparison pages from G2 review themes and AI research

Scope: Write down the exact page types this batch may include: direct “A vs B,” “A alternatives,” category shortlists, or migration pages. For each type, define the intended reader, buying stage, primary conversion action, and the commercial reason the page deserves to exist.

Guardrails: Record non-negotiable guardrails: no invented competitor facts, no copied review language, no outdated pricing claims, and no absolute claims unless legal and product owners have approved them.

Approval: Assign one SEO owner, one product-marketing reviewer, and one final approver so disputed claims have a clear escalation path. Do not begin keyword research until the team agrees on these rules and on the maximum number of pages it can properly maintain.

Set up: Open Ahrefs Keywords Explorer and enter each competitor name with modifiers such as “vs,” “alternative,” “pricing,” “reviews,” “migration,” and “best for.” Export keyword, parent topic, country, monthly volume, keyword difficulty, traffic potential, current SERP features, and top-ranking URLs into one target matrix.

Configuration: Add business fields for strategic competitor tier, ICP overlap, product readiness, approved proof availability, and expected conversion action. Score every target from 1-5 for search intent, attainable ranking opportunity, commercial fit, evidence readiness, and maintenance risk, then calculate a weighted priority score. Manually inspect any target that scores highly only because of volume, because broad brand queries may have weak buyer intent.

Approval: Approve a first batch of 5-10 URLs and park the rest in a backlog with the reason they were deferred.

Scope: For each approved keyword, use Ahrefs SERP Overview and Thruuu to capture the top ten results, page type, title, publish or update date, word-count range, heading pattern, comparison-table presence, FAQ coverage, and dominant source type. Use Perplexity in a separate research thread to summarize the apparent decision questions buyers are trying to answer, but keep its summary separate from verified SERP observations.

Action: Save the exact query, country, device assumption, and research date so the baseline can be recreated later. Label each observation as direct SERP evidence, tool-derived metric, or analyst interpretation.

Next action: Flag keywords where third-party review sites dominate because those pages will require stronger neutrality, first-party proof, and information gain.

Approval: End the step with a one-sentence search-intent statement for every target and have the SEO owner approve it.

Set up: Create one evidence ledger row for every factual statement the page may make. Use fields for page slug, claim ID, claim text, claim type, subject company, source URL, source title, source date, capture date, evidence excerpt or summary, owner, approval status, expiry date, and allowed wording.

Configuration: Add your own product proof in the same structure, including screenshots, documentation, customer evidence, implementation data, and approved pricing or packaging references.

Action: Mark competitor pricing, security, compliance, and roadmap information as high-risk and require an authoritative primary source plus a human approver.

Guardrails: Separate facts from positioning interpretations so the writer cannot accidentally present an inference as verified truth. Stop the workflow for any page whose core differentiation cannot be supported by current evidence.

Set up: Open the relevant G2 profile and collect a balanced sample across rating levels, company sizes, roles, industries, and review dates rather than selecting only negative comments. Assign each review a local review ID and record rating, reviewer role, company segment, review date, stated use case, and the review section where the theme appeared. Capture enough reviews to reach theme saturation, then note the point at which additional reviews stop producing new evaluation criteria.

Guardrails: Do not paste personal information, gated content, or long verbatim review passages into the working file.

Action: Keep direct quotations out of the page unless the quote is permitted and independently approved.

QA: Have the product-marketing reviewer inspect the sample for cherry-picking before analysis begins.

Set up: Open Claude and start a new project or chat dedicated to one competitor so evidence does not leak across pages.

Action: Paste the approved prompt into the chat, then attach the review table with local review IDs and include your sampling notes. Require Claude to return a structured table keyed to review IDs, with frequency, segment differences, confidence, and contradictory evidence for every theme.

Next action: Save the response as the “review-theme analysis” for that competitor and copy only the structured output into the research workspace.

QA: Reject any theme that cannot point back to at least two review IDs unless it is explicitly labeled as an isolated signal. Have a human reviewer compare a sample of the output against the source reviews before accepting the analysis.

Set up: Open a fresh Claude chat and paste the review-theme analysis, the approved search-intent statement, and the target ICP.

Action: Ask Claude to convert the themes into decision criteria such as setup effort, admin burden, support responsiveness, integration depth, reporting flexibility, total cost, or fit by company size. Require a matrix showing which criteria are strongly evidenced, weakly evidenced, or absent for each competitor.

Next action: Save the matrix as the buyer-criteria map and do not yet add your product’s claims. Review the criteria with sales or customer-facing stakeholders to catch dimensions buyers discuss in deals but public reviews underrepresent.

Approval: Approve only criteria that are relevant to the query and useful for an actual purchasing decision.

Scope: Compare the buyer-criteria map with the approved first-party proof inventory.

Action: For every criterion, identify the exact proof your product can support, the approved wording, the proof owner, and any evidence gap.

Next action: Use Claude only to organize the mapping; the product-marketing reviewer must decide whether each proposed claim is fair and material. Mark rows as supported, partially supported, unsupported, or not applicable, and prohibit unsupported rows from appearing in the comparison table.

Output: Record where the competitor may honestly be a better fit so the page can remain credible. If the page lacks enough supported differentiation, change the page angle or defer publication instead of manufacturing contrast.

Action: Use the approved intent, buyer criteria, and proof mapping to define one page thesis rather than trying to win every comparison dimension. Specify the primary audience, situation that triggers the comparison, most important decision criteria, fair-fit boundary, and conversion action.

Build: Build a decision path that moves from “what are you comparing” to “which option fits which situation” to “what proof supports the choice.” Record secondary audiences that the page will not optimize for so later edits do not dilute the thesis.

Approval: Have SEO and product marketing jointly approve the thesis before the brief is written. If they disagree, resolve the evidence or audience issue instead of blending two incompatible page angles.

Set up: Open Claude and paste the approved page thesis, SERP baseline, buyer-criteria map, proof matrix, and internal-link inventory into the prompt.

Action: Require a section-by-section brief with search purpose, heading, reader question, evidence IDs, required visual or table, internal links, CTA role, and assigned owner. Include title-tag and meta-description options, canonical recommendation, schema candidates, and the questions that must be answered above the fold.

Next action: Save the result as the controlled brief and give it a version number. The SEO owner checks intent coverage and information gain, while product marketing checks positioning and proof.

Approval: No drafting should start until both reviewers mark the brief approved.

Build: Draft the “best fit” summary and comparison table before writing the narrative because these elements force explicit decisions.

Action: Use one row per approved buyer criterion and include concise wording for your product, the competitor, evidence IDs, segment caveats, and a neutral interpretation. Keep “not verified” distinct from “feature absent,” and never treat missing public information as a weakness.

Set up: Add a short section explaining when each option may be the better fit. Product marketing verifies every row against the evidence ledger and blocks any row that overstates a difference.

Output: Save the approved table as a reusable structured object so future refreshes can update facts without rewriting the entire page.

Set up: Open Claude in the same controlled project as the approved brief and paste the copy prompt, brief version, approved table, and evidence ledger.

Build: Generate one section at a time: introduction, evaluation guidance, best-fit sections, key differences, migration or switching considerations, proof, and CTA. Require every factual sentence to carry an evidence ID in drafting notes, even though those IDs may be removed from public copy later.

Action: Save each section separately with status fields for drafted, SEO reviewed, product reviewed, legal reviewed, and final.

QA: Reject generic filler that does not answer a buyer decision question or add information beyond the current SERP. Assemble the page only after all blocks pass their assigned review.

Action: Use AlsoAsked to export related questions for the exact comparison query and country. Group questions into buying fit, implementation, migration, pricing, support, security, and category education, then remove duplicates and low-intent curiosities.

Set up: Open Claude and provide the approved evidence ledger plus the selected questions, requiring concise answers and a source or evidence ID for any factual claim.

Action: Mark questions that need product, legal, security, or pricing review and keep them out of public copy until approved.

Guardrails: Choose only FAQs that materially help the buyer decide; do not add filler merely to lengthen the page. Give the SEO owner the final question set and a separate list of candidates that may qualify for structured data under the site’s implementation rules.

Scope: Give the complete draft, comparison table, FAQ set, and evidence ledger to a reviewer who did not write the page.

Set up: Open Claude only as a first-pass checker and require a claim-by-claim risk table; do not treat its approval as legal approval.

QA: Verify that every competitor statement is current, supported, fairly worded, and distinguishable from opinion. Recheck pricing, package names, security, compliance, integration, and support claims against primary sources on the review date. Route high-risk items to the appropriate legal, security, finance, or product owner and record the disposition.

Approval: The page cannot move to publish-ready until every flagged claim is approved, rewritten, or removed.

Scope: Compare the approved draft against the original search-intent statement and SERP baseline.

QA: Verify title tag, H1, heading hierarchy, canonical behavior, indexability, internal links, image alt text, CTA destination, mobile table behavior, page speed, and any approved schema implementation. Check that the page offers clear information gain instead of merely matching competitor headings. Test every CTA and form using both business and personal email scenarios if routing differs.

Capture: Record the final URL, publish date, brief version, evidence-ledger version, owners, and next review date.

Output: Publish only when SEO, product marketing, and the final approver have signed off.

Capture: Record a pre-publish baseline for ranking position, impressions, clicks, CTR, organic sessions, CTA clicks, conversions, assisted pipeline signals, and branded-query movement where available.

Action: Use Ahrefs to monitor ranking and SERP changes, and review site analytics on a fixed 30-, 60-, and 90-day cadence.

Next action: Separate ranking problems, CTR problems, engagement problems, and conversion problems because each requires a different fix. Trigger an immediate evidence refresh when competitor pricing, packaging, branding, security, or product positioning changes.

Output: Log every test and revision against the page version so the team can see what changed and why. Retire, consolidate, or redirect pages that no longer have search demand, truthful differentiation, or an accountable owner.