If you have ever wired an LLM into a CMS, you have probably had this conversation with someone on the growth team: "Can we just generate the blog now? Will Google even rank it?"
It is a fair question, and the short answer is yes, AI-assisted content can rank. The longer answer is more useful to engineers, because it turns out ranking has very little to do with which tool produced the text and a lot to do with what the pipeline around the tool looks like. That is a systems problem, and systems problems are something developers are good at.
This post treats "content that ranks" like any other production workload: define the spec, identify the failure modes, then build guardrails so bad output never ships.
What Google actually evaluates
Let us start with the spec, because a lot of the online noise comes from people arguing about requirements that do not exist.
Back in February 2023, Google published guidance on AI-generated content. The gist: using AI does not give a page any special advantage, and it does not trigger a special penalty either. In Google's framing, it is just content.
What Google does evaluate is quality, the same way it always has:
- Is the page helpful and original?
- Does it demonstrate E-E-A-T (Experience, Expertise, Authoritativeness, Trust)?
- Or is it thin, duplicative or misleading?
Origin is not an input to that function. Quality is.
There is data that lines up with this. An Ahrefs analysis from 2025, covering more than 600,000 pages, found that 86.5% of pages in Google's top results contained at least some AI-assisted content, and the correlation between AI usage and ranking position was essentially zero. Put simply: Google is neither rewarding nor punishing the tool. It is grading the output.
The same dataset has a second number worth remembering. Only 4.6% of the pages analysed were pure AI, and those rarely reached position one. So "100% generated, zero human input" is technically able to rank, but it is not where the top results come from.
The five failure modes (think of them as bugs)
If the tool is neutral, why does so much AI content flop? Because teams treat it as a volume play: more posts, faster, less thinking. That produces a predictable set of defects. Here they are, written up like bug reports.
Bug 1: No original angle
Symptom: The article is a remix of the first page of search results.
Root cause: LLMs synthesise what already exists. If the prompt asks for "an article about X", you get the average of everything already written about X.
Why it matters: Content that only repackages known ideas gets deprioritised, and that pressure grows as AI Overviews answer generic questions directly in the results page.
Bug 2: Weak structure
Symptom: Long, rambling paragraphs, no clear hierarchy, headings that do not say anything.
Root cause: Nobody designed the information architecture before generating.
Fix direction: Headings, short paragraphs, lists and a logical flow. Readers and crawlers both parse structure.
Bug 3: Missing authorship and voice
Symptom: The text reads like nobody wrote it.
Root cause: No human experience was injected anywhere in the process.
Why it matters: Pieces from operators who add their own lessons tend to outrank generic ones.
Bug 4: Thin content
Symptom: 800 words that say almost nothing.
Root cause: Word count was the target, not insight.
Why it matters: Google's helpful content update is designed to push this kind of page down. Depth, real examples and actionable advice are what survive.
Bug 5: No E-E-A-T signals
Symptom: No clear author, no evidence of expertise, nothing that builds trust.
Root cause: The publishing template never asked for them.
Why it matters: Pages without these signals rank worse, regardless of how the words were produced.
Notice that none of these bugs is "an AI wrote it". Every one is a process defect.
A content pipeline that ships quality, not volume
Here is how I would structure the workflow if I were building it as a pipeline, using the five steps the original guidance recommends.
Stage 1: Strategy before tokens
Before generating anything, define the intent. Useful questions to answer up front:
- What do our users (for a marketplace: operators, buyers, sellers) actually need answered?
- What comparisons are they searching for?
- Which misconceptions do we need to correct?
Then connect that intent to subject-matter expertise, your internal linking plan and an AI-friendly structure. This is your spec. If you cannot write it down, do not hit "generate".
Stage 2: Humans write the core
Have someone with real experience write the opening, the main examples and the key insights. AI is then used to expand, refine and organise around that core. In other words, the human provides the signal and the model handles the formatting and filler that would otherwise eat an afternoon.
Stage 3: Structure as a deliberate step
Treat structure as a lint rule, not an afterthought: clear headings, short paragraphs, lists and white space.
Stage 4: Inject original data
Add case studies, examples from your own platform or business-specific numbers. Specific insights beat generic information every time, and they are the one thing a model cannot invent for you without lying.
Stage 5: Optimise for both search surfaces
Classic rankings and AI citations are related, but they are not the same thing. It helps to understand the difference between SEO tooling and AEO (answer engine optimisation) tooling. The good news is that the same hygiene, clear headings, real data and logical structure, helps on both fronts.
A pre-publish checklist you can actually enforce
You can turn the above into a simple gate in your CMS or PR template. Something like:
content_review:
intent_defined: true # which question does this answer?
original_angle: required # what do we say that the top 10 do not?
human_core: # intro, examples, insights by a named expert
author: "<name, role>"
structure:
max_paragraph_sentences: 4
descriptive_headings: true
original_evidence: ">= 1" # case study, platform example or own data
eeat:
byline: true
expertise_note: true
internal_links: ">= 2"
ai_role: "research | outline | expand | edit" # never "everything"
None of this is exotic. It is the same discipline you would apply to a code review: define what "done" means and refuse to merge until it is met.
Why this matters more for marketplaces
For marketplace and SaaS teams, content is often the most scalable growth lever available. It is cheaper than paid acquisition, it compounds over time and it builds authority. But it only compounds if it is strategic.
When competitors start publishing AI content at speed, the wrong response is to publish even more, even faster. The better response is to use AI to produce better content at scale. Teams that write deep, expert material about their own vertical tend to outrank teams pumping out thin AI-only articles.
If you run on a platform such as Sharetribe, your product is your moat. Content that explains how you are different, what you have learned and which real problems you solve is exactly the kind of material generic AI output cannot replicate. AI is the tool; your expertise is the advantage.
Content, SEO and engineering are one system
One last observation from the engineering side. Many teams split content, SEO and platform development into separate silos. The strongest growth usually comes from treating them as one system: a well-structured marketplace with clean category pages, sensible internal linking and authority-building content on top.
That is where developers can contribute far more than they think. Clean URL structures, fast pages, crawlable category templates and solid internal linking are engineering work, and they amplify every article the content team ships.
Takeaway
- AI-assisted content can rank. Google grades quality, not origin.
- Most AI content fails because of process bugs: no angle, weak structure, no voice, thin depth, no E-E-A-T.
- Fix the pipeline: strategy first, human core, deliberate structure, original evidence, optimise for both search and AI answers.
- For marketplaces, your domain expertise is the part a model cannot fake.
If you want to go deeper, the original article on whether AI content ranks on Google covers the strategic side in more detail. And if your team wants help connecting content, technical SEO and platform architecture, the SEO services team at Journeyhorizon works on exactly that intersection, alongside the broader marketplace work at Journeyhorizon.
How does your team gate AI-assisted content before it goes live? I would love to see other checklists in the comments.
Top comments (0)