--- language: en title: "Prompt for writing a comparison article" description: "Compare options across the same dimensions, cite sources for each attribute, use equivalent scopes, and tie recommendations to usage scenarios." canonical: "https://tropk.ai/en/resources/prompts/prompt-for-writing-a-comparison-article" last_modified: "2026-10-12T16:14:34.576Z" --- # Prompt for writing a comparison article Reviewed on 2026-10-06. ## Prompt to adapt and use Support the choice between named options using equivalent criteria and current evidence. ### Prepare before you begin COMPANY AND BUSINESS company: {empresa} domain: {dominio} confirmed_offering: {produtos_servicos_capacidades_limites_e_fontes} business_context: {mercado_regiao_diferenciais_comprovados_e_restricoes} AUDIENCE AND OBJECTIVE audience: {perfil_tarefa_conhecimento_inicial_e_restricoes} main_question: {pergunta_do_leitor} reader_outcome: {o_que_a_pessoa_deve_conseguir_decidir_ou_fazer} editorial_objective: {objetivo} search_intent_and_stage: {intencao_de_busca_e_etapa_da_decisao} VOICE AND EDITORIAL CONTEXT voice: {tom_vocabulario_exemplos_de_estilo_e_termos_a_evitar} topic_and_entities: {tema_keyword_principal_termos_e_entidades} inventory: {paginas_existentes_e_decisao_criar_atualizar_ou_consolidar} contribution: {exemplo_matriz_metodo_ou_dado_proprio_autorizado} cta: {proxima_acao_pertinente_e_destino_confirmado_ou_nao_aplicavel} EVIDENCE AND ACCESS sources: {id_titulo_url_autor_data_de_publicacao_consulta_e_trechos_relevantes} authorized_data: {dados_definicoes_periodo_metodo_e_permissoes_ou_nao_disponivel} internal_links: {url_publicada_assunto_e_funcao_do_link} research_access: {ferramentas_realmente_habilitadas_e_acessos_autorizados_ou_sem_navegacao} restrictions: {confidencialidade_direitos_prazos_mercado_e_escopo} options: {nomes_exatos_fornecedores_e_elegibilidade} criteria: {dimensoes_unidades_e_relevancia_para_o_publico} weights: {metodo_justificado_ou_comparacao_nao_ponderada} matrix: {opcao_criterio_valor_plano_mercado_fonte_data_e_incerteza} scope: {data_versao_regiao_e_planos_comparados} tests: {protocolo_e_resultados_reais_ou_nao_realizados} commercial_interest: {relacao_da_empresa_com_as_opcoes} ### Delivery format A. Markdown comparison with method, table, analysis by criterion, scenarios, and limitations. B. Evidence matrix by option and attribute. C. Verification questions and a record of tests actually conducted. ### Review criteria • The options serve the task and use the same criteria. • Plans, units, dates, and terms are comparable, or the difference is explicit. • Every material attribute has a source, and unconfirmed fields remain marked as such. • Recommendations specify scenarios and constraints, without invented scores or weights. • The method, tests conducted, and commercial interest are described accurately. ### Modelo Act as a research and writing assistant for the company provided. Write in Brazilian Portuguese, clearly and in the brand’s voice, for the audience’s task. Produce a draft for human review. Use the full context below, not just the topic. MODULE 1 — CONTEXT TO COMPLETE COMPANY AND BUSINESS company: {empresa} domain: {dominio} confirmed_offering: {produtos_servicos_capacidades_limites_e_fontes} business_context: {mercado_regiao_diferenciais_comprovados_e_restricoes} AUDIENCE AND OBJECTIVE audience: {perfil_tarefa_conhecimento_inicial_e_restricoes} main_question: {pergunta_do_leitor} reader_outcome: {o_que_a_pessoa_deve_conseguir_decidir_ou_fazer} editorial_objective: {objetivo} search_intent_and_stage: {intencao_de_busca_e_etapa_da_decisao} VOICE AND EDITORIAL CONTEXT voice: {tom_vocabulario_exemplos_de_estilo_e_termos_a_evitar} topic_and_entities: {tema_keyword_principal_termos_e_entidades} inventory: {paginas_existentes_e_decisao_criar_atualizar_ou_consolidar} contribution: {exemplo_matriz_metodo_ou_dado_proprio_autorizado} cta: {proxima_acao_pertinente_e_destino_confirmado_ou_nao_aplicavel} EVIDENCE AND ACCESS sources: {id_titulo_url_autor_data_de_publicacao_consulta_e_trechos_relevantes} authorized_data: {dados_definicoes_periodo_metodo_e_permissoes_ou_nao_disponivel} internal_links: {url_publicada_assunto_e_funcao_do_link} research_access: {ferramentas_realmente_habilitadas_e_acessos_autorizados_ou_sem_navegacao} restrictions: {confidencialidade_direitos_prazos_mercado_e_escopo} MODULE 2 — FORMAT INPUTS Format: comparison article Objective: Support the choice between named options using equivalent criteria and current evidence. options: {nomes_exatos_fornecedores_e_elegibilidade} criteria: {dimensoes_unidades_e_relevancia_para_o_publico} weights: {metodo_justificado_ou_comparacao_nao_ponderada} matrix: {opcao_criterio_valor_plano_mercado_fonte_data_e_incerteza} scope: {data_versao_regiao_e_planos_comparados} tests: {protocolo_e_resultados_reais_ou_nao_realizados} commercial_interest: {relacao_da_empresa_com_as_opcoes} MODULE 3 — EVIDENCE AND LIMITS Read the supplied texts. If research is enabled and authorized, consult relevant primary sources and record the pages actually read. Without browsing, use the attachments and return a list of outstanding checks; do not claim to have opened URLs. A competitor’s page can guide research but does not substantiate all its claims. Do not invent features, prices, data, customers, experiences, tests, credentials, testimonials, quotations, or links. Distinguish documented facts, observations, interpretations, and hypothetical examples. Missing information is neither zero nor proof that a feature does not exist. Preserve dates, versions, regions, plans, units, denominators, and material conditions. Derived numbers require calculations checked with a real tool or results already validated in the supplied inputs. Without that, mark verification as pending; do not claim to have recalculated. Do not disclose private data. Use names, images, and statements only with authorization. Do not copy the wording or the complete sequence of a reference. Preserve human authorship and indicate which validations require the subject-matter owner. MODULE 4 — FORMAT PRODUCTION 1. Define the decision, eligibility, and audience scenarios. 2. Normalize attributes by unit, date, market, and plan; preserve unconfirmed status. 3. Build the matrix with sources for each material cell and explain the method. 4. Analyze the criteria with equivalent depth across options, distinguishing documentation, testing, and attributed opinions. 5. Recommend options by scenario, disclose interests, and provide questions to confirm before purchase. Do not invent scores or a universal winner. MODULE 5 — SEO, AEO, AND GEO Use an H1 that matches the question and body; organize H2s/H3s around reader decisions. Start sections with direct answers when helpful. Each passage should retain enough entities, context, and conditions to be understood on its own. Use terms naturally, without fixed keyword density or artificial repetition. Cite the specific source near the fact it supports; preserve the supplied URL. Suggest internal links only to confirmed, useful destinations. Do not force an FAQ, word count, or special schema on every format. Do not promise rankings or citations. Do not add authorship, review, publication date, testing, or updates that did not occur. The CTA must match the stated offering and action, without a pitch unrelated to the task. MODULE 6 — OUTPUT AND REVIEW A. Markdown comparison with method, table, analysis by criterion, scenarios, and limitations. B. Evidence matrix by option and attribute. C. Verification questions and a record of tests actually conducted. Also provide proposed H1, SEO title, description, and slug; a claim map with source, section, and limitation; links used; and outstanding items with impact and next check. Keep this material separate from the publishable body. Do not publish or claim the draft is approved. Review against the criteria below and mark each as met, needs correction, or not applicable, with a reason: 1. The options serve the task and use the same criteria. 2. Plans, units, dates, and terms are comparable, or the difference is explicit. 3. Every material attribute has a source, and unconfirmed fields remain marked as such. 4. Recommendations specify scenarios and constraints, without invented scores or weights. 5. The method, tests conducted, and commercial interest are described accurately. If essential evidence for the central promise is missing, return only the safe portion and say which input prevents a conclusion. Do not fill the gap with a likely conclusion. ## When to use this prompt Use a comparison when readers need to choose between two or more identified options that serve a similar purpose. Start with the decision, not a desire to declare a winner. Record the audience profile, need, budget if known, and constraints. Criteria such as coverage, implementation, and how an option is used help only when it is clear why they affect this audience’s choice. This format requires evaluating the options on the same basis. Comparing one company’s basic plan with another’s full plan without explaining the difference leads to a weak conclusion. Define the market, region, period, version, and commercial terms. If the offerings serve different purposes, separate the scenarios or reconsider the format. An organized table cannot make incomparable data comparable. [SEO](/en/resources/glossary/seo) organizes the page for people searching; [AEO](/en/resources/glossary/aeo) focuses on clear answers; [GEO](/en/resources/glossary/geo) addresses discovery in generative responses. In this resource, all three are treated as writing, research, and review practices: address the task, name entities, preserve context, and make evidence verifiable. A [prompt](/en/resources/glossary/prompt) helps prepare a draft. It does not verify facts or guarantee rankings, recommendations, or citations by an AI. ## Prepare the inputs before writing Build a matrix with criteria in rows and options in columns. For each material attribute, provide the value, unit, plan, conditions, URL, and date checked. Prices need a currency, billing period, and terms. If a field is unconfirmed, keep it marked that way. “We didn’t find it” does not mean “it isn’t offered.” Include weights only if the brief provides a method and a reason for using them. Include official documentation for features and terms, as well as your own records if you conducted a real test. User opinions or independent reviews may be attributed, but should not be treated as universal proof of performance. Disclose the company’s commercial relationship with the options. For original contributions, use a task-based matrix, evaluation questions, or a protocol actually carried out; do not invent scores. ## How to produce and review the content The prompt organizes the method, table, and interpretation by scenario. Give the option profiles similar depth, and turn differences into practical implications: a limitation may matter to a small team and be irrelevant in another use case. A recommendation can favor one option, provided it states for whom, under what conditions, and on what evidence. The conclusion need not pretend to be commercially neutral, but it must remain factually accurate. During review, work through one dimension at a time across all options. Check units, dates, and conditions for every option. Look for claims of absence, exclusivity, superiority, and results that go beyond the source. Vendor-published case studies must remain attributed and should not be compared as if they followed the same protocol. When a decisive attribute is missing, ask the question the buyer needs to confirm. ## Example of a completed brief Example of a completed brief: compare two solutions for a team that needs to track questions and review sources. Criteria: documented scope, organization of observations, and [terms of use](/en/terms). Scope: same market and date. Original contribution: a task-based matrix and questions for a demo. Without an in-house test, the text describes documentation; it does not claim measured speed, a final score, or performance the team has not observed. ## How to use the full prompt Copy the full prompt and replace the fields in braces before sending it. Attach the source content as well as the URLs: a model without browsing cannot verify a link on its own. Use “not available” when a data point is missing and explain why. Generate a first version, review the outstanding items, and correct the inputs before requesting a revision. Keep the public-facing text separate from research notes. ## Check the result - The options serve the task and use the same criteria. - Plans, units, dates, and terms are comparable, or the difference is made explicit. - Every material attribute has a source, and unconfirmed fields remain marked as such. - Recommendations specify scenarios and constraints, without invented scores or weights. - The method, tests conducted, and commercial interest are described accurately. ## Sources and references - [Create helpful, reliable, people-first content]() — Google Search Central - [AI features and your website]() — Google Search Central - [Guidance on content created with generative AI]() — Google Search Central ## Continue a leitura - [Prompt for writing an alternatives article]() - [Prompt for writing a list article]() - [Prompt for writing a research study or benchmark]() - [AI discovery strategy]() - [Resources]() ## About the author [Luiz Gardelin]() Luiz Gardelin is the founder of Tropk.ai and works in GEO/AEO, product vision, strategy, and growth, helping brands understand and improve their presence in AI responses. --- [All pages in Markdown]()