AI Content Strategies That Backfire

Over the previous few years, I’ve watched AI content material creation instruments quickly acquire adoption throughout the Search engine marketing/GEO business. These instruments provide the promise of leveraging AI to automate content material creation, cut back headcount, reduce prices, and scale output.

As somebody who has spent the final decade serving to corporations get better from Google algorithm updates, my spidey senses began tingling the minute I heard the pitches for a lot of of those instruments. Even earlier than AI was a part of the dialog, Google already had a long history of lowering the visibility of automated content material in its search outcomes.

Regardless of latest developments within the high quality of AI outputs, I’ve remained skeptical that publishing AI-generated or AI-assisted content at scale can drive sustained efficiency in Google’s search outcomes. That is very true now, given how Google updated its ranking systems in recent years particularly to demote overly optimized, Search engine marketing-driven content material.

Over the previous a number of months, I’ve been monitoring greater than 220 web sites that had been publicly recognized, both by themselves or by their AI content material distributors, as prospects of assorted AI content material creation, automation, and scaling platforms. These instruments totally write articles, help with writing them, or use AI automations and workflows to help content material creation. Many of those instruments additionally now concentrate on driving visibility, mentions, and citations in AI search responses (AEO/GEO).

I wished to investigate what occurs after the claims of huge wins.

A constant sample emerged throughout the 220+ websites I’ve been monitoring, and I imagine it’s regarding sufficient to be value writing about: it really works, till it doesn’t.

Beneath, I’ll share a few of the traits I’m observing, plus a wide range of frequent Search engine marketing/GEO approaches I imagine could also be inflicting declines in natural search (and consequently, AI search) visibility. As a reminder, what’s harmful for Search engine marketing will also be harmful for AI search, largely due to RAG.

Methodology & Disclaimers

Earlier than we dive in, it’s vital to set the stage with my strategy and supply some vital disclaimers.

This evaluation relies on third-party Search engine marketing measurement information: natural site visitors estimates and natural web page depend time sequence information from Ahrefs, corroborated in opposition to the Sistrix Visibility Index information to substantiate broader visibility patterns. Prime-traffic URLs had been recognized utilizing Ahrefs’ top-pages export. The place I describe URL patterns or proportion adjustments, I’m quoting immediately from these third-party instruments as of Might 2026.

The dataset covers greater than 220 consumer domains tracked throughout the publicly revealed customer-stories pages of over a dozen AI content material platforms. For a lot of of those websites, I narrowed the evaluation to a particular subfolder the place the AI-assisted content material had been revealed, both recognized immediately within the case research itself or inferred from a pointy enhance in new pages across the time of the case research’s publication.

The evaluation, conclusions, and proposals all through this piece mirror my very own skilled opinions primarily based on greater than a decade of serving to corporations get better from Google algorithm updates. Different Search engine marketing/GEO practitioners could disagree with my findings and approaches, and particular person websites and techniques will all the time have their very own context.

3 Vital Disclaimers About This Information:

First, these are third-party estimates, not first-party analytics. They’re well-validated instruments within the Search engine marketing business, however they aren’t good measurements of natural search efficiency.

Second, the site visitors declines described right here might mirror many elements, together with however not restricted to algorithmic changes by Google, on-site adjustments by the positioning operators themselves, off-site aggressive dynamics, model adjustments, acquisitions, seasonality, and adjustments to inside web site structure. I’m not asserting that any AI content material software immediately brought about any site visitors final result described on this piece. I’m describing a correlation noticed throughout many listed websites that share comparable content material patterns and natural site visitors trajectories.

Third, distributors and particular domains are intentionally not named right here. The sample is the story, not the precise actors. Any resemblance to a selected firm, vendor, or case research is incidental to the broader sample described.

What The Information Exhibits: Fast Development Earlier than A Steep Decline

If there may be one factor the information makes clear, it’s this: scaling content material manufacturing with AI isn’t a low-risk technique for natural search. It may possibly produce actual short-term good points in each Search engine marketing and AI search (LLMs use search engines), however throughout this dataset, these good points have hardly ever held. In lots of instances, the eventual loss has exceeded the preliminary peak.

Throughout the group of 220+ websites and subfolders I analyzed:

• 54% misplaced 30% or extra of their peak natural site visitors.
• 39% misplaced 50% or extra.
• 22% misplaced 75% or extra.

Inside these declines, a recurring trajectory seems: a fast progress in natural pages over six to 12 months; an natural site visitors peak inside roughly three to 6 months of the content material peak; after which a steep decline in site visitors that erases a lot of the acquire (and often drops beneath the prior baseline) inside the following 12 months.

Most of those site visitors drops befell after the case research had been revealed (which additionally makes me ponder whether the case research themselves might be contributing to the declines). Within the instance beneath, the case research was revealed in January 2025, indicated by the the black star beneath:

I’m additionally repeatedly monitoring adjustments to natural web page progress and natural site visitors to those websites and subfolders over time. Trying on the up to date information, a considerable variety of these manufacturers seem to have considerably lowered their content material footprints in 2025 and 2026, usually eradicating, redirecting, or 410’ing most of the identical pages featured as success tales in revealed case research. This might clarify the latest drop in pages (yellow line) proven within the above screenshot (and probably, the corresponding enhance in natural search site visitors).

In lots of instances, these case research stay revealed to this present day, however the pages they reference don’t.

The Acquainted Rank & Tank Playbook

When a web site begins seeing site visitors drops on account of sitewide content material high quality points, it’s hardly ever a mild decline. As Glenn Gabe refers to it, a greater label could be “Mount AI”: steep progress, adopted by a equally formed drop-off in natural site visitors, as soon as Google’s methods have gathered enough signals to identify what is going on.

Beneath are a number of examples of case research websites that used AI to scale content material creation and noticed huge drops in natural site visitors after their case research had been revealed:

This sample is constant throughout industries, together with cybersecurity, journey, advertising and marketing, SaaS, healthcare, B2B providers, crypto, and client items, and it exhibits up throughout distributors.

The form of the road within the chart is just like trajectories we’ve got seen amongst many websites affected by Google’s algorithm updates lately. It’s the identical boom-bust cycle the Search engine marketing business has watched repeatedly in numerous varieties, accelerated this time by the velocity at which AI instruments have enabled web site house owners to scale content material.

The Search engine marketing Trade Simply Went By means of This

What is difficult to overstate is simply how lately the Search engine marketing business watched a near-identical cycle play out. Many SEOs and web site house owners are nonetheless licking their wounds from a brutal spherical of Google updates and new spam insurance policies that obliterated many websites’ site visitors just a few years again.

In September 2023, Google launched the Helpful Content Update, probably the most aggressive crackdown it had accomplished in years in opposition to content material that, according to its announcement, “feels prefer it was created for search engines like google and yahoo as a substitute of individuals.”

Roughly six months later, in March 2024, it adopted up with the longest core replace in Google’s historical past, which Google states was designed to “reduce unhelpful, unoriginal content in search results by 45%.” Throughout two consecutive replace cycles, Google’s acknowledged goal was the identical factor: content material produced at scale, no matter whether or not the manufacturing technique was human, AI, or a mix of each.

Alongside the March 2024 replace, Google formalized a brand new spam coverage known as “Scaled Content Abuse,” explicitly naming the apply it was working to suppress: producing many pages to control search rankings, no matter authorship.

The Search engine marketing business continues to be working by means of the collateral injury from these updates, together with important losses for a lot of small publishers, a few of whom had been publishing authentic, human-written content material however used extreme Search engine marketing frameworks that the updates seemingly flagged. The casualty checklist additionally included some publishers who had partnered with advert networks and different rising instruments providing AI content material creation and scaling as a service.

Many sites affected by the HCU haven’t recovered to this day, regardless of their huge efforts. I spent important time in 2024 working and talking with many web site house owners attempting to dig themselves out of that gap.

Having spent a whole lot of hours analyzing and presenting about these two main updates, I can say that the content material I’m seeing revealed with many of those new AI instruments usually appears to be like and feels lots like the precise sort of content material that was wiped off the map with these 2023 and 2024 Google updates.

8 Recurring Content material Patterns That Are Dangerous For Search engine marketing And AI Search

So, what varieties of content material am I seeing revealed by corporations utilizing AI instruments to construct articles that I imagine are in the end dangerous for Search engine marketing? I imagine the reply lies in web page templates that purpose to affect Search engine marketing rankings, AI search responses, and/or citations in AI search, however are highly formulaic and easily repeatable by competitors.

What begins as a real strategy to attempt to construct useful content material (and rating a point out/quotation) finally ends up being an simply detectable footprint by Google when sufficient websites are publishing comparable pages, and the index turns into flooded with tens or a whole lot of 1000’s of those comparable pages, which is simpler than ever to do utilizing AI.

That is precisely what Google means when it talks about writing for search engines like google and yahoo, not people.

Reviewing top-traffic URLs throughout the declining domains, eight distinct content material templates seem repeatedly. Most websites seeing declines within the evaluation use some mixture of at the very least three or 4. Essentially the most aggressive ones use all eight. Usually, affected websites even have a whole lot or 1000’s of those articles, which amplifies the issue and customarily results in steeper site visitors losses.

1. Comparability Pages At Scale

Sample: /weblog/[product-A]-vs-[product-B] revealed at scale throughout most affordable head-to-head matchups in a class. Noticed throughout the dataset for product-vs-product pairings, framework-vs-framework pairings, and, in at the very least one case, concept-vs-concept pairings unrelated to the writer’s precise enterprise.

2. The “What Is X” Glossary

Single-term, single-question pages designed to be cited by AI engines. Sample: /sources/what-is-[term] or /glossary/[term]. Noticed throughout the dataset, together with programmatic glossaries scaled throughout a number of languages from a single supply template. Scaling translations with AI and with out human assessment also can often result in sitewide content material high quality points.

3. The “Finest [X] For [Y]” Listicle

Essentially the most acquainted AI-content template, with origins within the affiliate-content period. This sample was noticed throughout the dataset in each broad-category and narrow-niche variants.

4. The Self-Promotional Listicle

A variant of No. 3 by which the writer is itself a competitor within the class being ranked, and often lists itself because the No. 1 greatest amongst rivals. These pages usually lack actual proof that the corporate genuinely examined the entire rivals within the checklist, which is recommended by Google for review pages.

I wrote about this “listicle” page template causing SEO/GEO issues in February 2026, when I discovered that many corporations publishing dozens, a whole lot, and even 1000’s of self-promotional listicles noticed excessive site visitors drops starting on the identical day (roughly Jan. 21, 2026). This sample was noticed throughout a number of websites within the dataset, most aggressively in B2B providers.

5. The Competitor-Vs-Options Web page

Sample: /weblog/[competitor-brand]-alternatives, or, within the extra programmatic type, devoted touchdown pages constructed for each named competitor in a class. This strategy was noticed extensively throughout the dataset, together with one case the place the majority of a web site’s high site visitors pages had been devoted to particular person competitor model names.

6. Programmatic Location And Language Scaling

This is among the oldest tips within the Search engine marketing guide, and one which I’ve seen websites get in hassle for with algorithm updates for at the very least 10 years. The strategy: Use one template multiplied throughout each geography or language a search engine will index, with little or no distinctive content material per native touchdown web page.

In lots of instances, the corporate publishing these pages usually doesn’t have actual brick-and-mortar areas in every of the neighborhood/metropolis/state pages they’re concentrating on.

This web page sort was noticed throughout the dataset together with state-by-state content material, country-by-country service pages, and the multilingual programmatic glossaries described above.

7. The FAQ Farm

Every web page solutions precisely one query. Sample: /faq/[full-question]. Designed for extraction by AI engines: a transparent query within the URL, the reply within the first paragraph, bullet factors within the physique, schema markup on the backside.

The issue? This strategy creates a number of low-quality content material and baggage for the positioning when applied at scale. Scaling FAQs was additionally noticed extensively throughout the dataset, together with in industries the place the templated tone was a noticeable mismatch with the writer’s model context.

Here’s a screenshot of my March 2024 Amsive article advising against the same exact thing:

It’s additionally value noting that simply final week, Google introduced it was deprecating FAQ Wealthy Outcomes, which I believe might have something to do with this new influx of FAQ schema aimed toward attempting to earn citations and mentions in AI search.

8. Off-Subject Content material Printed At Scale

Publishing off-topic content material, with no obvious connection to the writer’s precise enterprise, at excessive volumes, is among the quickest methods to get in hassle with search engine algorithms. This was additionally an enormous drawback in the course of the Useful Content material Replace and March 2024 Core Updates, when many websites had been experimenting with publishing off-topic content material, like humorous quotes, jokes, child names, horoscopes, and different high-volume articles that weren’t truly topically related for the writer.

This technique was used throughout a number of websites within the dataset, together with items on leisure subjects on a providers platform, lists of names and jokes, social-media memes on B2B web sites, and historic or biographical content material on business-focused websites.

The Late January 2026 Unconfirmed Google Replace

A secondary sample seems within the information round late January 2026: a wave of sites with explicitly GEO-optimized, self-promotional listicles, plus different dangerous Search engine marketing/GEO approaches, noticed natural site visitors declines between 40% and 95% over the January-April 2026 window.

Google didn’t announce or verify an replace by identify in January 2026, however at the very least 40 websites I recognized noticed a unfavourable pattern starting round Jan. 20, 2026. In lots of instances, the affect was remoted to the corporate’s weblog or different subfolder containing a number of new Search engine marketing-driven content material. My evaluation discovered that a few of these corporations had been scaling dozens, a whole lot, and even 1000’s of those self-promoting listicles, by which they named their very own firm the No. 1 greatest when in comparison with rivals.

I believe this adjustment on Google’s finish was simply the beginning of Google (and sure the LLM suppliers constructing on high of search) starting to demote this kind of content material in search outcomes, and it seems that the affect was better than simply the listicles themselves. For affected websites, your entire weblog or subfolder containing these articles usually additionally noticed declines. In different instances, the affect was carried over throughout the total area.

How To Use AI Content material Instruments Safely

I do imagine there’s a method to make use of AI content material instruments safely, and a method for these instruments to help the creation of high-quality content material. The instruments themselves should not the issue, however the implementation might be. I imagine these instruments needs to be used and overseen by skilled Search engine marketing professionals who perceive the panorama of content material approaches that Google has grown extraordinarily refined at penalizing and demoting over the previous 10+ years. The issue usually stems from a “set it and overlook it” strategy, or when the purpose is to scale as many pages as shortly as doable with out human assessment.

Utilizing AI content material instruments for analysis, group, content material briefs, pulling in proprietary firm information and insights, and extra might be invaluable for dashing up the content material creation course of. However when articles are merely revealed “for Search engine marketing/GEO” with out consideration of the dangers concerned with search engine rating methods, the well-intentioned content material can truly backfire for each Search engine marketing and AI search.

To carry out nicely, I like to recommend that any AI-assisted content material ought to nonetheless show E-E-A-T, add authentic or distinctive info above and past what is obtainable by competing pages (information gain), and take into account being clear about the usage of AI to create the content material (which is really helpful by Google).

The Backside Line

If there may be one takeaway from monitoring these 220+ websites over the previous a number of months, it’s that the playbooks being bought as “AI-first Search engine marketing” or “GEO-optimized content material at scale” look remarkably just like the playbooks that bought websites flattened by the Useful Content material Replace and the March 2024 Core Replace. The packaging is new, however the sample isn’t.

Throughout the dataset, the manufacturers nonetheless rising are usually those whose content material doesn’t match the eight templates above. Many manufacturers that scaled into these templates are those now eradicating pages, redirecting subfolders, and taking different steps to attempt to mitigate latest losses in site visitors.

When you’re at the moment evaluating an AI content material vendor, or working a program in-house, listed below are just a few sensible questions I feel are value asking earlier than publishing one other web page:

• Does this web page truly exist as a result of an actual buyer or reader wants it, or as a result of a search engine or LLM may cite it?
• May a competitor publish a near-identical model of this web page tomorrow utilizing the identical immediate?
• Would I be snug if Google, a journalist, or my very own prospects noticed the total checklist of URLs on this subfolder?
• Is the article inherently biased, and if that’s the case, is the web page clear with customers about these biases?
• Is there any first-party information, experience, or authentic perspective on this web page that isn’t out there on the primary ten outcomes already rating for the question?

None of this implies AI content material instruments are unusable. They are often genuinely helpful for analysis, briefs, inside information synthesis, and accelerating workflows where a human expert is still in the loop. The difficulty begins when the purpose turns into quantity, or when the individuals closest to the content material cease reviewing what goes out the door.

The Search engine marketing business has already lived by means of this cycle as soon as in the previous couple of years. The websites that got here out of it greatest had been those that prioritized high quality, originality, and topical focus over scale. I count on the identical to be true of this cycle, and I’ll preserve monitoring the information because it performs out.

Extra Assets:

This post was originally published on Lily Ray NYC Substack.

Featured Image: Stokkete/Shutterstock