Nathan Gotch's Process for Stopping Category Decay

Nathan Gotch has a name for the slow slide that hits sites in Google and AI answers: category decay. Instead of watching one page lose clicks, he watches a whole topic across every search surface and acts before the drop shows up in traffic.

I went through his latest video and pulled out the full process, step by step, including the parts you can do with free tools.

What category decay is

Content decay is the old idea. A page ranks well, then six to twelve months later its impressions, rankings and clicks start to fall. We have a full guide on how to find and revive aging pages if that is the problem in front of you.

Gotch's point is that a single page is now the wrong unit. A category is every page, query and AI prompt around one topic. His example from his own company is "best AEO tools." He wants to know whether the brand is winning that whole category in traditional search and in AI answers, and he wants to see dips early.

He splits the causes into two lists. Things you control: content quality, entity coverage, technical health, content variety, backlinks, freshness, internal links, distribution, expanding into new subtopics and brand mentions. Things you don't: algorithm updates, AI model changes, seasonality, competitors catching up, falling demand for the topic, search intent shifts, SERP layout changes like AI Overviews and more ads, click loss to AI answers and platform policies.

The second list never stops working against you, which is why he says SEO is never one and done. Let a category slide too far and the work to bring it back is far bigger than the work to hold it.

Step 1: Pick one category and list its queries and prompts

Choose one topic that makes you money. Write down the queries people type into Google for it, then turn the same topics into prompts someone might ask ChatGPT or Gemini.

Gotch's example is a plastic surgeon in New York City. The category is liposuction, so the list covers every type of liposuction and every kind of intent, each tracked on its own. That lets you see the overall category score and then the weak subtopics inside it.

He is upfront that prompt tracking is controversial, because nobody knows exactly what people type into AI tools yet. You can still know the general topics where you need to be visible, and that is enough to build the list.

Milestone: one category, split into subcategories, with a query list and a prompt list for each.

Step 2: Track the category on every surface

Pull the category's numbers from Google Search Console, Google Analytics and Bing Webmaster Tools. Filter by the pages and queries in your category so you see the topic as one line over time, not as hundreds of separate rows.

Then check the other surfaces that matter for your business:

• AI answers: ChatGPT, Gemini, Perplexity and Google's AI features, run against your prompt list.
• Local: map pack and local grid results if you serve a location.
• Traditional search: Google, Bing, DuckDuckGo and Brave.
• Social search: mainly YouTube, LinkedIn and X.

The point is to compare yourself against the top competitor in each subcategory. If you lift the weak subtopics, Gotch says the whole category rises with them.

Milestone: a monthly snapshot of the category per surface, with your position next to your strongest competitor.

Step 3: Watch the leading metrics

Traffic is a lagging number. Gotch watches three things that move first, all counted for the category:

1. How many new assets you published.
2. How many existing assets you updated or upgraded.
3. How many brand mentions and links you earned.

When he investigates a category that is declining, he says he almost always finds all three trending down at the same time. Publishing slows, upgrades stop and mentions dry up, and then the traffic follows.

Milestone: those three counts logged every month next to the traffic line.

Step 4: Build your knowledge base

Before writing anything, Gotch builds what he calls a search brainiac: one place that holds everything a writer, human or AI, needs about the business. It has five layers.

• Data: Search Console, Analytics, Google Business Profile if you are local, YouTube, Microsoft Clarity and your rank tracking.
• Company: products, services, FAQs, policies and brand voice.
• Expertise: interviews with the subject matter experts. He does this first on every campaign. Sometimes that is the founder, and sometimes it is someone on the team who knows the details better than the founder.
• Search intelligence: your swipe file, keyword lists and content inventory.
• Execution: a record of every page created, every page optimized and every technical change. His team uses GitHub for this. Without a log you can't check whether your actions did anything.

Milestone: a folder or project where all five layers live, starting with at least one recorded expert interview.

Step 5: Build a template library from the top five

For each keyword in the category, look at the top five competitors and write down how their content is structured: sections, order, formats, what they include near the top. Save that as a template.

Gotch is clear that this is modelling a structure that already works in search and in AI citations, not copying anyone's text. Repeat it for every keyword in the category and you end up with a library of proven structures.

Milestone: one saved structure template per keyword in the category.

Step 6: Research Google results and AI citations together

The old method was to take the top 10, sometimes 25 or 30, Google results for a keyword and extract the entities and subtopics they cover. Gotch says that still works for Google.

His update is to add the pages that AI answers cite for the same topic. Those sources give you a wider set of entities and subtopics, and when his team started doing this, their pages began to perform in AI citations as well as in traditional search.

Focus on the competitors with the most search coverage, since they have already shown what works. For a couple of pages, a quick bookmarklet or a small script you write with an AI assistant is fine. He warns that doing this by hand for 30 to 50 pages per keyword does not scale.

Milestone: a list of entities and subtopics per keyword, pulled from both Google's top results and the AI citation set.

Step 7: Cover the fan-out questions

When an AI tool answers a question, it runs a set of related searches in the background to build its answer. Gotch's way to cover those is to collect the People Also Ask questions for the keyword and answer them on the page.

He covers them on the main page first. If one of those sub-questions underperforms, he splits it off into its own page later.

Milestone: every relevant People Also Ask question answered on the page, with a note of any that might deserve their own page.

Step 8: Write brief 1.0, then 2.0 with your purple cow

Brief 1.0 comes from the research: structure, entities, subtopics and fan-out questions. Gotch says you could stop there and still beat most people.

Brief 2.0 is what keeps a page ranking. You add the knowledge base from step 4, your brand voice, and something only you have. He calls it the purple cow.

His example is his best SEO books article. He bought and read 27 SEO books and wrote from that experience. He says it has ranked for four or five years, through site changes and domain redirects. When it started to slip because he stopped watching it, a quick refresh with current entities sent it straight back up. He says lifts like that often show up within 24 to 48 hours.

Milestone: a 2.0 brief that names the first-hand material the page will include.

Step 9: Choose human or AI by competition, then draft

Gotch's study found no statistical advantage for human or AI content on raw performance. What he has seen anecdotally, after testing both since 2021, is that good human-written pages tend to hold their rankings longer and need fewer refreshes.

He also shows a case study that ranked for years and was written by AI from start to finish. It worked because the input was a first-party study his team ran. The raw material was unique, so the output was too.

His rule is simple. If the keyword is very competitive, he writes it by hand. For medium and low competition, he leads with AI.

Milestone: a first draft, written the way the competition level calls for.

Step 10: Fill the gaps and run the on-page basics

Compare the draft against the entities and subtopics from step 6 and fill whatever is missing. In Gotch's study, the pages with the most visibility were the ones that covered the topic most completely.

Then the on-page basics, which he says work just as well for AI as for Google:

• Topic in the URL slug. 86% of the URLs in his study had an exact or partial match.
• Topic in the title tag and the meta description.
• Topic in the H1.
• Topic in the first 100 words. The evidence was thinner here, but the outliers that beat everyone else had it.
• A variant of the topic in at least one H2, ideally the first.

We covered his full pre-publish checklist in our article on on-page SEO checks, so I won't repeat it here. Finish with an AI pass for editing and quality checks.

Milestone: no missing subtopics against your research, and the topic placed in all six spots.

Step 11: Check the page against his design rules

Gotch calls these his laws of intelligent content design. Run every page against them:

• Match the intent. A plumber in a specific town needs a service page built to convert. Don't build an informational page for a commercial search, or the other way round.
• Give it a goal. Every page should push the reader deeper into the funnel or produce a lead.
• Make it fast. He says there is no excuse for slow sites anymore.
• Make it skimmable. People still read websites. Build for them and you are usually fine for agents and algorithms too.
• Use visuals that add something. Relevant images and video break up the page. Decorative stock photos and filler images just take up space.
• Use HTML. 92% of the AI-cited pages in his study were HTML.

Milestone: the page passes all six before it goes live.

Step 12: Do the other 80% after you publish

Gotch's least popular slide: publishing is only about 20% of the work. After the page is live:

1. Annotate and track the page and the category over time.
2. Add internal links from your power pages, your strongest URLs, to the new one for an early lift.
3. Repurpose the topic on YouTube, LinkedIn and anywhere else you publish, so you own more of the results page.
4. Prioritize the top retrieval sources. Look at what the AI answers cite, check whether you appear there, and reach out to get mentioned.
5. Keep building links. Your site is the part of the AI's inputs you control, and AI tools use search results during retrieval.
6. Let it cook for 30 days without changing the page.
7. Use Search Console to find cluster topics. After 30 days, look at the new queries the page shows up for, even at position 75, and turn the promising ones into supporting pages. Our guide to topic cluster strategy explains how to structure those.

Then go back to step 2 and keep watching the category.

Milestone: after 30 days, a list of new supporting pages pulled from Search Console, added to the plan for the category.

Automate it with Distribb

Twelve steps per category, repeated every month, is a lot for a founder or a small team. Gotch shows a tool doing most of it at the end of his video. Here is how the same process maps to Distribb, step by step, and what stays with you.

• Steps 1 to 3, tracking and leading metrics: Distribb reads your Search Console clicks and rankings for every page and tracks when AI answers cite you. The content calendar shows every article planned for the category on a date, so you can see your publishing pace without a separate log.
• Steps 5 to 10, research and writing: Distribb does the keyword research, plans the articles on the calendar, writes them and publishes them to your site, linked to your existing pages. Your writing instructions, banned phrases and brand voice apply to every article.
• The refresh: Optimizations finds pages that slip and fixes them. If you work from Claude or ChatGPT, you can ask which pages lost clicks this month and have the worst one rewritten on the same URL with its internal links kept.
• Step 12, links and mentions: the backlink exchange adds 3 to 5 contextual links a month from real businesses in your niche. Link Outreach reverse engineers the sources ChatGPT cites and drafts the pitches to get you mentioned there.

What stays with you is the part Gotch says makes pages stick: the expert interview and the first-hand material from step 4 and step 8. Put that into your project's business details and writing instructions, and the rest runs on autopilot.

Pro is $97 a month with 30 articles. Accelerator is $495 a month with 90 articles, done-for-you link outreach campaigns, videos and Medium articles. Both start with a free trial.

Start your Free Trial

Watch the video

The video is "Avoid This Deadly AI SEO Mistake" on Nathan Gotch's channel, just under 30 minutes. The slides on what you can and can't control are worth pausing on.

FAQ

What is category decay in SEO?

Category decay is a drop in visibility across a whole topic, measured in traditional search and in AI answers together. Nathan Gotch uses the term to move attention away from single pages. You track every page, query and prompt in one category, such as "best AEO tools," and act when the category line starts to fall.

What is the difference between content decay and category decay?

Content decay is one page losing impressions, rankings and clicks over six to twelve months. Category decay looks at all the pages and prompts around a topic on every surface, including AI answers. It also tracks leading signals like publishing pace, updates and new mentions, which usually drop before traffic does.

How do you know if a category is declining?

Look at the category's combined numbers in Search Console, Analytics and Bing Webmaster Tools, plus your visibility in AI answers for its prompts. Then check three leading metrics: new assets published, assets updated and mentions or links earned. Gotch says a decline usually shows all three falling together.

Should I write SEO content with AI or by hand?

Gotch's rule depends on competition. For very competitive keywords he writes by hand, and for medium or low competition he leads with AI. His study found no statistical difference in raw performance, though his experience is that good human-written pages hold rankings longer. AI pages built on first-party data performed well for him.

How long should you wait before changing a new page?

Gotch waits 30 days. After that, he checks Search Console for the queries the page has started to show up for, even at low positions like 75, and uses them to plan supporting pages. Changing the page sooner makes it hard to tell what is working.

Pick your first category today

You don't need to track your whole site to start. Pick the one category that brings in the most revenue, write its query and prompt list, and log the three leading metrics this month. If you want the publishing, refreshes and links handled for you while you watch the numbers, see how Distribb's AI SEO agent works from Claude or ChatGPT.