How to audit a year of content with AI in one sitting

I published 67 posts before I ever looked back at any of them. When I finally did an audit, I found four posts covering essentially the same topic in slightly different ways, a guide recommending a tool that had shut down eight months earlier, and one post quietly getting a third of my search traffic that I’d completely forgotten writing.

The audit took a full week, mostly spent reading my own back catalogue. The version I run now takes an afternoon, and the difference is entirely in what I hand to AI.

Why auditing beats publishing

The uncomfortable arithmetic: a post that already ranks on page two can often be moved to page one with two hours of work. A new post starts at zero and takes months to find out whether it’ll rank at all.

For anyone with more than about thirty pieces, updating is the highest-return content activity available, and almost nobody does it because it isn’t as satisfying as making something new.

Step 1: Get everything in one table

You can’t audit what you can’t see at once. Build one row per piece with:

• URL and title
• Publication date and last update
• Traffic over the last 90 days
• Where it ranks, if you track that
• Primary topic
• Whether it has a call to action, and which

Export what you can from your analytics rather than typing it. Twenty minutes of setup, and this table is the whole audit, everything else is annotation.

I keep mine in Notion because I want it next to the content calendar it feeds, and because the previous year’s audit sitting in the same database is genuinely useful context. A spreadsheet works identically. My Notion AI review covers whether the AI layer adds anything, and Notion vs Obsidian for solopreneurs if you’re picking a home for this.

Step 2: Sort into four buckets

Every piece gets exactly one label. Resist inventing a fifth.

Update. Gets traffic or ranks close to something good, but the content has aged. This is your priority list and where nearly all the value is.

Merge. Overlaps substantially with another post. Fold the good part into the stronger piece, redirect the weaker URL.

Delete. Genuinely wrong, embarrassing, or about something that no longer exists, with no traffic and nothing worth saving. Should be a very short list.

Leave alone. Everything else, which will be most of it. Posts that are fine, doing their job, and don’t need your attention. Giving yourself permission to skip these is what makes the audit finishable.

Step 3: Let AI do the reading

This is the part that used to take a week.

Overlap detection. Paste titles and descriptions for everything and ask which pieces substantially overlap, with reasoning. It catches pairs you’ve forgotten about, and it’s genuinely better at this than you are because it has no attachment to any of them.

Staleness check. Feed it a post and ask what’s likely out of date: tools mentioned, prices, statistics, references to years. Fast, and it flags the specific lines rather than a vague verdict.

Gap analysis. Give it the full list of titles and ask what an audience interested in this topic would expect to find that’s missing. This is where next quarter’s calendar comes from, and it pairs well with turning your comments into content ideas, which sources the same answer from your actual audience rather than from inference.

Structural review. For each update candidate, ask what questions the post fails to answer. That list becomes the new sections.

The prompt discipline that matters: ask for analysis of what’s there, not for suggestions of what would be nice. Otherwise you get a generic list of best practices that applies to any website.

Step 4: Do the updates properly

A real update is not a date change. What actually moves the needle:

• Add the missing section. Usually the biggest single win. The post was 80% of an answer, and someone bounced looking for the other 20%.
• Fix what’s factually stale. Dead tools, changed pricing, superseded advice.
• Improve the opening. Most old posts start with throat-clearing you’ve since learned to cut.
• Add internal links to things you’ve published since. Your old posts don’t know your new posts exist, and this is free.
• Update the call to action. Old posts frequently point at offers that no longer exist.

Then, and only then, change the date. Doing it in the other order is the thing that gives updating a bad name.

Step 5: The merges

Merging is fiddly and worth doing properly. Pick the stronger URL, usually the one with more traffic or better links. Move the genuinely useful parts of the weaker piece into it, rather than pasting the whole thing. Then redirect the old URL to the new one permanently.

Don’t skip the redirect. An orphaned URL throws away whatever the old post had earned and produces a dead link for anyone who bookmarked it.

Step 6: Write the follow-up plan

The audit’s real output isn’t the updates, it’s the list of what to make next, and it’s better than anything you’d brainstorm because it comes from evidence.

Three lists fall out naturally: posts that performed well and deserve a deeper sequel, gaps where your archive has an obvious hole, and topics where you now have a genuinely different opinion than you did.

Those go into the calendar with dates. How to build a 30-day content calendar with AI covers turning that list into a schedule, and a simple AI SEO workflow for solo bloggers is the process for the pieces that need keyword work.

My take

Block one afternoon, twice a year. Build the table, sort into four buckets, let AI read everything and flag overlap and staleness, then update your top ten and merge the obvious duplicates.

The reason to do it isn’t tidiness. It’s that the highest-return content work available to you is almost certainly sitting in something you already published, half-finished, ranking on page two, one missing section away from being genuinely good.

And be generous with the leave-alone bucket. The temptation in an audit is to find something wrong with everything, which turns an afternoon into a month and produces a lot of pointless fiddling. Most of your archive is fine. Find the ten pieces that aren’t, fix those properly, and go back to making new things.