Ahrefs States Every Article Utilizes AI, Defining "Slop" as Content Lacking Human Understanding and Effort Transfer

Ahrefs has revealed that every article published on their platform incorporates AI in some capacity, emphasizing that the extent of AI usage is not the sole determinant of content quality. The company defines "AI slop" not by stylistic quirks like excessive em dashes or particular phrasing, but as content that is published without sufficient human understanding, judgment, evidence, or original contribution to warrant a reader’s attention. Essentially, AI slop shifts the burden of effort from the creator to the reader, allowing publishers to bypass crucial steps like investigation, verification, opinion development, and relevance assessment, leaving the reader to discern trustworthiness and understanding.

To counter this phenomenon, Ahrefs outlines a four-pronged approach to leveraging AI effectively while maintaining content integrity. The core principle is to preserve the practices that historically made content valuable, even as AI takes on a larger role in the creation process.

1. Prioritize Human Work Before Drafting:
The company stresses that with AI significantly reducing the cost of prose generation, more effort must be directed upstream to the foundational stages of content creation. This involves meticulously deciding what topics are worthy of exploration, how to structure an article, and what unique contributions can be made. Before engaging AI for drafting, writers are encouraged to establish a clear premise that addresses key questions about the topic’s significance, target audience, and potential contributions. AI can assist in answering these questions by mapping search results, challenging angles, identifying counterarguments, or highlighting missing evidence, but these critical decisions must remain human-led.

Crucially, AI needs substantial, original raw material to work with. Providing it with only publicly available internet data will likely result in generic content. Ahrefs advocates for incorporating unique assets such as interviews, internal knowledge bases, proprietary data, real-world demonstrations, accounts of failed experiments, and specific examples derived from the company’s own work. The concept of a "Source of Truth" is highlighted, exemplified by a tool built by a colleague, Mateusz, which serves as a searchable library for information relied upon by his agent and himself for content creation. This repository includes facts and statistics, explanations, product details, and how-to guides.

A significant practical shift for individual writers involves prioritizing verbal communication over typing during the initial ideation phase. Instead of self-editing while typing, which can inadvertently filter out valuable nuances like uncertainty, caveats, opinions, and nascent connections, writers are encouraged to "talk it out." Dictation tools like Wispr Flow are recommended for this purpose, allowing writers to freely express their thoughts, essentially creating an unedited "podcast episode" with themselves. This process offloads a wealth of raw material for AI to process. Alternatively, writers can prompt AI to interview them, identify gaps in their knowledge, extract potential claims, and propose outlines, thereby transforming AI from an inventor of substance into an organizer and interrogator of human-provided material.

2. Establish Decision Points and Opportunities for Judgment:
AI’s proficiency in embedding decisions within polished prose can obscure the critical choices made during content generation. A single, comprehensive prompt can simultaneously dictate research direction, angle, structure, claims, examples, and tone, presenting these decisions as a finalized article. To counteract this, Ahrefs advocates for a pipeline approach that mirrors human editorial workflows, rather than generating a single, opaque output file. This involves saving distinct stages of the process – research, identification of content gaps, outlining, and drafting – separately. This allows for inspection at any juncture, enabling the correction of problematic outputs or instructions and restarting from an acceptable point.

While a complex pipeline is not always necessary, the principle of breaking down work into stages with human judgment points in between remains vital. These stages might include initial brainstorming without drafting, followed by outline generation, and then drafting, with a decision point after each phase. At each "gate," writers must exercise judgment, which could involve requesting more research, deleting sections, altering the angle, or even abandoning the article. This practice is essential because the ease with which AI generates polished content can create a "sunk cost fallacy," where individuals feel compelled to refine existing output rather than questioning its fundamental necessity or direction. AI’s ability to rapidly produce large volumes of text exacerbates this issue. Even when using AI for a single article, breaking the process into distinct stages—brainstorming, outlining, drafting—with iterative refinement at each step, including asking AI to "steelman" arguments, is recommended. The focus should be on making better decisions, not merely increasing word count. Failing to do so risks transforming writers into editors before they have fully engaged as thinkers.

3. Allocate Saved Time to Enhancing Content Quality:
Ahrefs differentiates itself by focusing on how AI-driven efficiency can be reinvested in creating superior content, rather than solely in increasing output volume. While acknowledging that AI is used to automate tedious tasks and save time, enabling more frequent publishing and updating of articles (as demonstrated by their Data Refresh Hub), the company emphasizes that volume is not the only viable return on efficiency.

Historically, the constraints on content creation extended beyond drafting and publishing to the incorporation of valuable elements that were difficult to justify or invest in, such as simple data analysis requiring a data scientist, the creation of free tools needing a developer, or interactive articles demanding design and engineering expertise. Large research projects also presented a significant manual workload for individual content marketers. AI demonstrably lowers these barriers, empowering writers to perform data analysis, prototype tools, build interactive elements, improve article user interfaces, or automate aspects of research projects. While these capabilities may not achieve perfection without specialist intervention when stakes are high, they significantly reduce the threshold for experimentation.

This is illustrated by the transformation of a basic TablePress embed into a richer interface for exploring data, such as the most-cited domains in AI Mode. It also enables content marketers to create quizzes, free tools, or data visualizations without the lengthy wait for developer availability, exemplified by their free LLMs.txt generator. The key distinction Ahrefs emphasizes is whether AI is used to merely reduce the work involved in each article or to undertake tasks that were previously financially or logistically prohibitive. The critical question is not how many more articles AI allows a company to publish, but rather what new and previously impossible elements can now be integrated into an article. The capacity for scaling must be tempered with restraint; the objective is not to turn every writer into a content factory, but to expand individual writer capabilities without compromising the standards for published content.

4. Assign Ownership and a Second Human Reviewer to Every Article:
Even when AI executes a significant portion of the content creation process, a specific individual must ultimately own the final product. The person overseeing the AI-driven workflow needs sufficient topic expertise to validate claims, correct misinformation, explain the provenance of evidence, and ultimately decide whether they are comfortable attaching their name to the published work. This is why individuals like Ryan, Director of Content Marketing at Ahrefs, do not mass-publish articles, despite having an AI pipeline. As the editor of all content, he reads every word published on the Ahrefs blog. Furthermore, as an industry thought leader, Ryan possesses the critical capacity to challenge premises, question evidence, identify generic sections, and assess whether an article genuinely fulfills its promise to the reader.

A "human in the loop" is only effective if that human actively engages in critical review rather than merely rubber-stamping AI-generated output. This requires the human reviewer to possess the necessary knowledge, authority, and a willingness to reject subpar content. Consequently, even for articles that are heavily AI-generated or assisted, Ahrefs ensures sufficient human hours—from both the writer and the editor—are invested to guarantee the content meets publication standards.

Final Thoughts:
Ahrefs asserts that AI-generated content is not inherently "slop." Their research on how Google treats AI content indicates that heavily AI-generated pages can achieve high rankings. However, the company observed a general decline in search performance as AI usage increased, attributing this not to Google’s explicit punishment of AI content, but to companies using AI to circumvent the demanding work of content creation, leading to the publication of more average-quality material.

The fundamental issue, according to Ahrefs, is not AI itself, but the abdication of responsibility. AI has democratized content execution, freeing up time that should be dedicated to the judgment, evidence gathering, experiential insights, and original ideas that elevate an article. Therefore, the paramount question is not the quantity or origin of words generated by AI, but whether a human has truly understood, judged, verified, and endorsed the content before its publication. If no individual genuinely takes ownership of the final outcome, the result is deemed "slop."