A report on "How to Generate SEO Articles in Bulk From a Keyword List"

The graded article is 1,023 words, written by DURBUN from one keyword. Open it below to read what the models read. The report took 14.6 seconds and cost 607 credits.

Scores by axis

Each axis is the average of the three models. Where they disagreed the spread is shown, because a wide spread usually means the article is ambiguous rather than bad.

Search intent

6.7

Writing quality

6

Topical coverage

6

Information gain

5.7

Evidence and credibility

5.7

Structure

6

Spam safety

6.3

Strongest: Search intent, Spam safety. Weakest: Information gain, Evidence and credibility, Writing quality.

What each model said

GPT-5.6 Luna, 7/10. This is a practical, reasonably clear workflow that directly addresses bulk article generation and includes useful operational details such as batch limits, review steps, publishing controls, and credit costs. It is more useful than a generic AI-content overview, but it is heavily centered on DURBUN and leaves important parts of the workflow underdeveloped, especially keyword-list preparation, clustering methods, quality-control criteria, and avoiding large-scale search-content risks. The FAQ also adds repetitive API-limit guidance without much relevance to the core query.

GPT-5 Nano, 7.4/10. Overall solid how-to guidance for bulk SEO article generation using a specific tool. It is well-structured and actionable, but leans heavily on product promotion without independent validation, which slightly lowers credibility. It would be valuable as a companion or case-study page but not as a sole top-ranking resource without external context.

Gemini 3.8 Flash, 3/10. This article hijacks a broad, tool-agnostic query and turns it into a narrow onboarding guide for a single proprietary tool (DURBUN). It fails to cover the broader landscape of bulk generation workflows—such as Python/OpenAI scripts, Google Sheets automations, or competing SaaS platforms—and degenerates into repetitive, bolted-on FAQ spam about API rate limits that were never introduced in the body text.

What it got right

• The workflow is immediately actionable: "Build the list from real searches, give each line one search intent, merge lines that would compete for the same query, choose shared settings, review the finished articles, and publish at a steady pace." This gives the reader a sensible end-to-end process rather than only describing a tool.
• It identifies cannibalization as a real bulk-publishing risk: "When two terms would lead to nearly identical pages, merge them and keep the stronger phrase as the primary keyword." That is more valuable than simply telling readers to upload a keyword list.
• The article includes concrete operational constraints and costs: "A run takes 2 to 50 lines," "DURBUN writes up to 5 articles at once," and "1,000 credits cost $0.05." Specific limits make the workflow easier to plan.
• The review guidance correctly emphasizes editorial control: "Read every finished article for usefulness, search intent, factual support, natural keyword use, readable titles, and repeated or thin sections." This prevents the page from presenting bulk generation as a publish-without-review process.
• "Bulk Generate in DURBUN is a practical tool for this workflow."
• "DURBUN writes up to 5 articles at once, shows finished articles out of the total with a timer, and keeps completed articles if you stop the run."
• "One credit equals one output word, and 1,000 credits cost $0.05."
• "For a first run, prepare five keywords, generate them in one batch, review the drafts in the dashboard, and publish them as WordPress drafts."
• Provides practical, sound clustering advice before running generation: 'When two terms would lead to nearly identical pages, merge them and keep the stronger phrase as the primary keyword.'
• Gives exact pricing and operational metrics for the featured tool rather than vague assertions: 'One credit equals one output word, and 1,000 credits cost $0.05. AI images use 1,000, 2,000, or 6,000 credits each, while stock photos are free.'

What it wants fixed

• The article is too product-specific for the broader query. Much of the answer is devoted to "Bulk Generate in DURBUN" and its templates, languages, credit system, and WordPress features. Add a tool-agnostic process first—keyword export cleanup, clustering, input formatting, generation, QA, internal linking, and publishing—then present DURBUN as one implementation.
• Keyword-list preparation is asserted rather than explained. The instruction "Start with real search data rather than a raw export" does not say how to collect, normalize, deduplicate, cluster, or prioritize the terms. Add a worked example showing a raw list becoming grouped article inputs, including search intent, primary keyword, secondary terms, and the decision to merge or split pages.
• The quality-control section is too general for bulk publishing. "Check that the primary keyword matches the page" and "provides information beyond the keyword itself" are useful principles but not a repeatable checklist. Specify checks for factual sourcing, originality, title and heading fit, competing URLs, internal links, author or product claims, thin sections, and whether the article actually satisfies the dominant search format.
• The API-focused FAQ is repetitive and weakly connected to the query. "What should I do if my batch exceeds an API limit?" and "When is a separate request better than one large batch?" largely answer the same question, while the article does not establish that readers are using an API. Replace the FAQ with a concise troubleshooting section covering duplicate pages, mismatched intent, poor drafts, failed jobs, and batches containing mixed formats.
• The credit guidance contains a minor clarity problem: "One credit equals one output word" is followed by "1,000 credits cost $0.05," but the article does not explain whether generation, revisions, or all output modes consume credits. State exactly what is billed and show one complete sample calculation for a batch.
• Promotional tone and product-only framing without independent validation or benchmarks.
• Reliance on a single vendor (DURBUN) without alternatives or generalizable best practices.
• Some claims are broad (e.g., live search results and checks for overstated claims) without empirical evidence or testing data.
• Extensive product-specific steps may reduce applicability to readers using different tools or platforms.
• Potential redundancy in sections (e.g., very similar guidance across different formats and steps) could be tightened for conciseness.
• The FAQ section abruptly pivots to developer API concurrency and polling limits that have zero relevance to the preceding UI-based dashboard tutorial: 'Split the batch into smaller requests when it exceeds an API’s concurrency or per-minute start limit.' Replace this with genuine user questions regarding content quality assurance, deduplication, or internal linking workflows.
• The FAQ repeats the exact same concept three times across three different questions ('exceeds an API limit', 'separate request better than one large batch', 'paced instead of simply divided'). Consolidate or cut these entirely.
• Completely ignores search intent by treating a general how-to query as single-vendor documentation: 'Bulk SEO article generation turns a prepared keyword list into complete drafts, and Bulk Generate in DURBUN is a practical tool for this workflow.' Introduce tool-agnostic methodologies (e.g., custom scripts, CSV imports, Make/Zapier pipelines, programmatic SEO templates) and evaluate multiple approaches.
• The writing style is monotonous and staccato, with repetitive short declarative sentences that read like machine-generated instructional text.

How to read the score

The grader keeps nine and above for pages it would rank above the current results and publish itself, and it puts most published writing on the web at six or seven. The verdict follows the average: 8.5 and above reads Would outrank the current results, 7 Competitive, worth publishing, 5.5 Publishable after the fixes below, and anything lower Not ready to publish. Every fix above is quoted from a model, so each one points at a passage you can change.