Content Decay: A Content Refresh Strategy That Wins Rankings Back
Content decay is a slow loss of clicks you can spot early and reverse. Here is how I find decaying pages, decide what to do with each one, and refresh them every quarter.

Content Strategy
You can use AI for SEO content without losing E-E-A-T if AI handles structure and first drafts while people supply what a model cannot: first-hand experience, original data, expert judgement and a named author who stands behind the piece. Google's guidance focuses on whether content is helpful and made for people, not on how it was produced. What it treats as spam is content mass-produced mainly to manipulate rankings.
At Garage Collective I lead an 8-person content, technical SEO and off-page team across a 12+ brand portfolio. LLM-drafted outlines and first drafts, refined with SME input, original data and practitioner detail, are what let us publish 10 to 12 long-form pieces a month without the work turning generic. This post is that workflow: the stages, who owns each one, and the checklist that decides whether a piece goes live.
It is worth getting this right, because most of the bad advice on AI content comes from misreading it.
Google's position, paraphrased from its Search Central guidance: it rewards helpful, reliable, people-first content regardless of whether it was written by a person, with AI assistance, or both. Using automation is not against the rules in itself. Using it, or anything else, mainly to manipulate search rankings is.
Since 2024, its spam policies have a name for the second case: scaled content abuse. That covers producing large numbers of pages mainly to rank, with little value for the reader. The policy applies however the pages are made, so a room of people churning out keyword variations is as much a problem as a script doing it.
Google's helpful content guidance also encourages you to think about "who, how and why": who created the content, how it was created, and why. The "why" is the one that matters most. If the answer is "to get traffic" rather than "to help the person searching", the production method is irrelevant.
E-E-A-T (experience, expertise, authoritativeness and trust) comes from Google's search quality rater guidelines. Raters do not set rankings directly, but the guidelines describe what Google's systems try to reward. The first E, experience, is where unedited AI content falls apart. A model has never run a campaign, used a product or sat in a sales call.
In my experience, the split is clear once you have done it for a while.
| Good use of AI | Keep it with people |
|---|---|
| Turning a brief and SERP notes into an outline | Deciding what the piece should argue |
| Listing questions a section should answer | Answering them from real experience |
| A first draft built from SME notes | Numbers, examples and claims that need to be true |
| Summarising call transcripts and support tickets | Choosing which insight is worth publishing |
| Title, meta and FAQ variations to choose from | Final voice, tone and opinion |
| Spotting gaps against a competitor outline | Judging whether the gap is worth filling |
The pattern: AI is good at structure and at turning material you already have into readable prose. It is bad at producing the material. If you ask it for facts, examples or statistics, it will give you plausible ones, and plausible is not the same as true.
This is the order we run it in. The most important choice is that SME input comes before the first draft. If you bolt expertise on at the end, you get a generic article with a few expert quotes stapled to it.
| Stage | Who | What happens | Output |
|---|---|---|---|
| 1. Brief | SEO strategist | Target query, intent, funnel stage, what ranks now, what we know that those pages do not, internal links, named SME | One-page brief |
| 2. Outline | AI, checked by writer | Model drafts an outline from the brief and SERP notes, with open questions marked for the SME | Outline with gaps flagged |
| 3. SME input | SME, run by writer | Short call or voice note answering the flagged questions. Real examples, data, opinions, mistakes they have seen | Transcript and notes |
| 4. First draft | AI | Model writes from the outline plus SME notes only. Missing facts are left as marked placeholders, never filled in | Rough draft |
| 5. Rewrite | Writer | Rebuilds the draft in the brand voice, adds original data, fills placeholders from sources, cuts filler | Writer's draft |
| 6. Expert review | SME | Checks accuracy and adds anything that is missing or wrong | Approved facts |
| 7. Edit and publish | Editor, then SEO | Voice, structure, checklist, on-page SEO, schema, internal links | Live page |
A weak brief produces a weak draft no matter who or what writes it. The line I care most about is "what do we know that the ranking pages do not". For a B2B software client like MoneyGence, that might be how accountants actually use a feature. For a D2C brand, it might be what customers ask before buying. If nobody can fill in that line, the piece is not ready to be written.
Two instructions make the biggest difference to AI drafts. First, give the model the brief, the SME transcript and your notes, and tell it to use only that material. Second, tell it to leave a visible placeholder such as [SME: example needed] or [SOURCE NEEDED] wherever it lacks information. A draft full of placeholders is useful. A draft where the model quietly made things up is dangerous, because it reads fine.
Experts are busy, so do not send them a blank page. Send the outline with specific questions. "What do people get wrong about this?", "What would you do first?" and "Can you give one real example?" get better material than "Any thoughts?". In my experience, a recorded conversation, transcribed and summarised by AI, is the most efficient format.
This is where E-E-A-T is actually added. Sources of original material that most companies already have:
One specific detail from real work does more for a page than another 500 words of general explanation.
Give each brand a short style sheet: tone, words to use, words to avoid, how to address the reader, and two or three example paragraphs that sound right. The editor checks every draft against it. AI drafts drift towards the same flat, slightly formal voice, so the edit is where a Kaayu Rituals piece and a Utkarsh Small Finance Bank piece stop sounding alike.
Free checklist
39 checks to get your brand cited in AI Overviews, ChatGPT, Perplexity and Gemini
Every piece passes this before it goes live. If one item fails, it goes back a stage.
The answer-first formatting and clean structure also make pieces easier for AI search to cite, which I go into in how to get cited in Google AI Overviews.
Nobody sets out to break the spam policy. Teams drift into it when volume targets outrun review. Watch for these:
If two or more of these sound familiar, cut volume and fix the review step before publishing anything else.
Judge AI-assisted pieces the same way you judge any content. In Search Console, check that new pages get indexed and start earning impressions for their target queries. In GA4, look at engagement and key events from organic landing pages. For SaaS, connect content to MQLs, SQLs and pipeline, as I describe in my post on B2B SaaS SEO that drives pipeline.
Then keep the older pieces working. Volume creates a large library, and a large library decays. A quarterly refresh cycle is the other half of this workflow: it is where we update facts, add new practitioner detail and catch any page that has slipped.
AI did not change what good content is. It changed how fast you can produce the parts that were never the hard part. The experience, the data and the judgement still have to come from people, and that is the part worth protecting.
Google does not penalise content for being AI-assisted. Its guidance rewards helpful, people-first content however it is produced. Content mass-produced mainly to manipulate rankings, whether by AI, people or both, falls under its scaled content abuse spam policy.
Free AI search audit
Send your site and our founder reviews it personally. You get a clear, prioritised read on what is holding your search growth back.
Keep reading
Talk to the founder
Tell us what you want to rank for and where growth has stalled. We will tell you straight whether search is the right lever, and what it is worth.
Abhishek DwivediFounder and Lead Strategist, AD Search Studio
