Discussion starter: a content SEO question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- Content SEO
A SaaS blog built on AI-assisted posts drops after a core update. What would convince you the content caused it, rather than a niche-wide shift?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
No single metric would convince me. What would convince me is a pattern where three things line up: comparable sites held their positions on the same queries while you lost yours, the results that replaced you are the same kind of page (other vendors’ or publishers’ blog posts, not forums, videos or tool pages), and your losses are heavier in the posts that look weakest on substance than in comparable posts that do not. If those three hold, “the content caused it” is the best explanation available. If competitors fell alongside you or the results changed type, you are looking at a niche or intent shift, whatever your production process was.
One reframe before starting: “AI-assisted” is not the hypothesis. Google’s guidance on generative AI content says the problem is using AI “to generate many pages without adding value for users”, which may violate the scaled content abuse policy. The testable question is whether the posts were unhelpful, generic or redundant, not whether a model helped write them. That matters because the evidence you need is about substance and competition, not about detection scores.
Rule out the cheap explanations first
- Confirm the timing. Google’s core updates documentation recommends waiting at least a full week after a core update completes before analysing your site. Check whether the drop started inside the rollout window (the March 2026 and May 2026 core updates both completed this year) or was a gradual slide that the update merely made visible. Google also documents smaller, unannounced core updates, so not every step change has a named cause.
- Separate ranking loss from click loss. If average position on your main queries held but clicks fell, the core update may not be the cause. Informational SaaS queries (“what is X”, “how to do Y”) are exactly where AI Overviews compete for clicks. Search Console’s generative AI performance reports show AI impressions only, not clicks, but they tell you whether AI surfaces grew on your pages over the same period.
- Check for technical and manual causes. Deploys, template changes, internal link changes, canonical or noindex mistakes, and the Manual actions report. A scaled content abuse manual action is a different problem from an algorithmic reassessment.
- Check reporting artefacts. A year-on-year comparison that straddles September 2025 is affected by Google dropping support for the num=100 parameter, which pushed reported impressions down for many sites. Use clicks and non-brand queries (the branded queries filter helps) rather than raw impressions.
The evidence that separates “our content” from “the niche”
Signal Points to your content or site Points to a niche-wide shift How to check Competitor visibility on your lost queries Comparable SaaS blogs held or gained Most comparable sites fell too Track 10 to 20 peer domains on your keyword set in a third-party visibility tool, same dates Who replaced you Other blog posts or guides answering the same intent, often with more specific detail Forums, video, marketplaces, product or template pages, or local-market domains Compare current results for your top lost queries with saved snapshots or your rank tracker history Shape of the loss on your site Broad across topics, or heaviest in the weakest posts Concentrated in one query type that changed for everyone Page-level and query-level change, grouped by topic and content type Country split Similar loss in every market Loss concentrated outside your home market Search Console country filter, before versus after Indexing Rising “Crawled - currently not indexed” for posts that used to be indexed No change in indexing Page indexing report, filtered to the blog directory The country row is worth checking closely this year. Aleyda Solis’s analysis of the May 2026 core update found shifts were driven by intent, market fit and source type, including UK results moving from US.com retailers to local-market domains. That retail finding is not about SaaS, so treat it as a hypothesis to check, not a finding about your niche. Still, a US SaaS blog that lost most of its UK and Australian clicks while US traffic held has a market-fit question, not necessarily a content-quality one.
The cohort comparison, and how it misleads
The obvious test is to compare posts produced with heavy AI assistance against posts written or rewritten by subject experts. If the AI-assisted cohort lost far more, that looks damning. It often is not, because the cohorts differ in other ways:
- AI-assisted posts are usually newer, with fewer links and less accumulated engagement.
- They were often produced to cover broad, generic informational queries at volume, which are the queries most exposed to AI Overviews and to forum and video results.
- Volume production tends to create overlap: three posts targeting slightly different phrasings of the same question, splitting signals.
So match before comparing. Pair posts by topic, query type, pre-update clicks, age and referring domains, then compare losses within pairs. If the AI-assisted post in each pair still lost clearly more, that is real evidence. If the gap disappears after matching, the difference was about topic and age, not production method.
Also compare against your human-written posts. Google has said its core ranking systems work mainly at page level but also use some site-wide signals, and its core updates page notes that “deleting the unhelpful content can help the good content on your site perform better”. That implies weak content can weigh on the whole site. So if your older, expert-written posts also dropped while competitors held, that does not clear the AI-assisted cohort. It is consistent with a site-level quality assessment where the weakest cohort is a likely contributor.
Competitors held, results pages unchanged in type?↓No: niche, intent or market shift is the leading explanation. Adapt formats and targeting; do not mass-rewrite yet↓Yes, and losses concentrate in weak, generic or overlapping posts: page-level content problem↓Yes, and losses are broad including strong posts: likely site-level quality assessment, with the weakest cohort as the prime suspectWhat the content audit should look for
Run a manual review on a sample of lost posts against the pages that now outrank them. This is where the case either firms up or falls apart. Look for:
- Anything a competitor or a general model could not have written: screenshots from your product, real configuration steps, data from your own usage, pricing or limits you know first-hand.
- Whether the post answers the question the searcher actually has, or restates the definition and pads around it.
- Factual errors, outdated feature names or invented capabilities. On a SaaS blog these damage trust with buyers even if they never affect rankings.
- Overlap: multiple posts competing for the same query group.
- Authorship: whether a named person with relevant experience stands behind the post. Google’s helpful content guidance asks whether bylines lead to background about the author.
Watch for this:
Running your posts through AI detectors tells you nothing useful about why they dropped. The detectors are unreliable, and Google’s stated test is value to users, not production method. A human-written generic post and an AI-assisted generic post fail for the same reason.
Check what the lost traffic was worth
Before deciding how much to invest, look at which lost pages produced trials, demo requests or product sign-ups. Scaled top-of-funnel posts often attract traffic with little buying intent. If the lost clicks were mostly from generic definitional queries that never produced pipeline, the right response may be to consolidate them and put effort into product-led content (integrations, use cases, comparisons, templates) rather than trying to win back every lost visit.
The experiment that settles it
Hypothetical example:
A SaaS blog has 600 posts, 420 of them AI-assisted and published in the last 18 months. After the core update, non-brand clicks fell 45%. Five peer blogs tracked in a visibility tool are roughly flat. The results for lost queries still show blog posts, mostly from those peers. Within matched pairs, AI-assisted posts lost more than expert-written ones, and the expert-written posts also fell, but less.
Here the evidence points to content, at both page and site level. The team picks 40 matched pairs of lost AI-assisted posts: one post in each pair is rewritten by a product specialist with first-hand material, or merged into a stronger post, while the other is left alone. Separately, the thinnest 100 posts with no conversions and no links are consolidated or removed.
- Primary metric: non-brand clicks and query coverage for treated versus control posts.
- Window: at least through the next core update. Google’s documentation says it “could take several months” for its systems to confirm a site is producing helpful content in the long term. Page-level gains can show sooner; site-level recovery usually waits for a later update.
- Record everything: use Search Console’s custom annotations to mark each batch of changes.
- What changes the decision: treated posts recover relative to controls, so content was a lever and the rewrite programme is justified. Neither group recovers while peers hold, so the issue is site-level and consolidation of the weak cohort matters more than rewriting. Both groups recover as the niche recovers, so it was a niche shift after all.
A practical position
I would accept “the content caused it” when peers held, the results page still rewards blog content, and matched comparisons show the weakest posts lost most. Short of that, I would call it a niche shift with a possible content contribution, fix the obvious weaknesses (overlap, errors, generic filler) because they are worth fixing anyway, and hold back on a full rewrite until the controlled comparison gives an answer.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.