Google’s “Thin Content” Crackdown May Be an AI Content Reckoning: What SMEs & Agencies Must Change Now
A recent thin-content manual action reported by SEOs hints at a bigger shift: Google may be treating large volumes of AI-generated responses as “little or no added value.” Here’s what changed, why it matters, and a practical playbook for keeping your content—and your AI—eligible to rank and get cited.
By Marius Dosinescu (AYSA.ai)
Google has used the phrase “thin content” for years. What’s new—and what should make every business owner, publisher, and agency pause—is how that label may be expanding in practice to include large volumes of AI-generated pages and AI-written forum replies that don’t add anything meaningfully new.
A recent report covered by Search Engine Journal describes a partial Manual Action for “Thin content with little or no added value” applied to a large forum section. The SEO community speculated that a high-volume AI bot replying to users may have been a trigger—especially if those replies were framed as “staff” answers and produced at scale.
This matters because “thin content” is no longer just an affiliate problem or a doorway-page problem. In a world where AI can generate an infinite number of plausible paragraphs, “thin” increasingly means interchangeable: content that looks like an answer but doesn’t carry real experience, real specificity, or real accountability.
Concise summary
- Manual actions still happen, and Google can apply them to a subsection of your site (like a forum path), not just the whole domain.
- AI-generated volume is a risk multiplier: if you publish thousands of AI answers that don’t add original value, you create a “thin content” footprint even if individual pages look fine.
- UGC sites and forums are exposed: the promise of a forum is lived experience. AI replies can violate user expectations and reduce perceived authenticity.
- The fix isn’t “ban AI”. The fix is to design an editorial and technical system that forces “added value,” enforces quality gates, and prunes what doesn’t deserve Indexing.
- AYSA fits as an execution system: monitor risk signals, generate specific fixes, ask for approval, and safely implement accepted changes across content, indexing rules, and Internal linking.
Table of contents
- What changed: “thin content” is being interpreted more broadly
- The real issue: “thin” often means “no added value,” not “short”
- Manual actions: why this isn’t an algorithmic shrug
- Why forums, knowledge bases, and UGC sites are in the blast radius
- A practical risk model: when AI content becomes “thin” in Google’s eyes
- Why this is happening now: AI search is forcing a higher bar
- A concrete SME scenario: the local clinic with “helpful” AI pages
- What agencies must rethink (before clients get hit)
- The 2026 thin-content prevention playbook (step-by-step)
- If you got a thin-content manual action: triage and recovery plan
- The AYSA approach: approved execution beats “publish and pray”
- What to do next
- Sources and further reading
What changed: “thin content” is being interpreted more broadly
The headline from SEJ isn’t “Google hates AI.” Google has repeatedly said (in various public communications over the years) that it’s not the method of creation that matters most—it’s whether the content is helpful and high quality. The problem is that AI makes it easy to produce content that resembles helpfulness while being structurally empty:
- Repackaged definitions
- Generic advice that could apply to anyone
- Answers written without lived experience
- Pages that exist mostly to capture queries rather than solve problems
The SEJ report describes a forum that received a partial manual action targeting a specific URL pattern. The manual action message: “Thin content with little or no added value.” The controversy is that forums naturally contain some low-effort posts—yet the action appeared to be applied broadly to a forum area with hundreds of thousands of threads.
SEOs speculated that one differentiator might be AI-generated replies posted at high volume, potentially as a “staff” account, which could change the perceived authenticity of the forum’s content. This is not confirmed as the cause, but it’s a plausible hypothesis—and it’s consistent with how “thin content” is commonly described: content that does not add original insight, research, or expertise.
For business leaders, the takeaway is simple: Google’s tolerance for low-value scale content is shrinking. AI makes scale trivial, so the bar for value has to move upward—or search results become unusable.
The real issue: “thin” often means “no added value,” not “short”
Many teams still interpret “thin content” as “short articles.” That’s outdated.
“Thin” is better understood as content that fails one or more of these tests:
- Originality test: Would a reader learn something here they wouldn’t learn from the first three results?
- Specificity test: Is the answer tied to a real context (location, product constraints, pricing realities, steps, edge cases)?
- Experience test: Is there evidence of real-world usage, mistakes made, photos taken, or decisions explained?
- Accountability test: Is there a responsible author/editorial process—or is it anonymous bulk output?
- Outcome test: Does the content actually help a user complete a task, choose a product, or avoid an error?
AI can help you pass these tests—but only if you use AI to augment real knowledge, not replace it. A forum answer that reads like a textbook excerpt might be “correct,” but it doesn’t deliver the reason users chose a forum in the first place: the nuances of human experience.
Manual actions: why this isn’t an algorithmic shrug
Algorithmic ranking changes are noisy. Manual actions are not. If a manual action is applied, it means a human reviewer (or a workflow initiated by human review) decided a section of your site violates guidelines or quality expectations strongly enough to warrant suppression until you fix it and request reconsideration.
The SEJ story highlights a partial manual action: one URL pattern was affected, not the entire domain. That detail matters operationally:
- Your “money pages” can stay fine while a blog, forum, or support section gets suppressed.
- Or the reverse: your ecommerce/category pages may rank, but your informational layer collapses—killing discovery and top-of-funnel demand.
- It forces prioritization: you must identify which segments of the site are quality liabilities.
In practice, partial actions can be more dangerous than sitewide penalties because they trick teams into believing “we’re okay,” until the suppressed area turns out to be the part that feeds your pipeline.
Why forums, knowledge bases, and UGC sites are in the blast radius
Forums, communities, and UGC platforms are one of the last “human” layers of the web. People come for context:
- “I tried this and it broke.”
- “This worked in Arizona but not in Michigan.”
- “If you’re on version X, do this instead.”
- “Here’s a photo of what the error looks like.”
Now imagine a forum thread where the top response is posted by an “AI staff” account and it restates generic information that can be found everywhere. That can become a quality signal problem in two ways:
- User expectation mismatch: the “forum value” is diluted; the thread becomes redundant with general web content.
- Scale effect: if the AI account posts tens of thousands of replies, you create a large, uniform footprint of similar answers—easy for quality systems to classify as low-value.
And it’s not just forums. Consider other common business assets:
- Support knowledge bases: AI-generated troubleshooting articles that don’t match real product behavior can look “helpful” but increase returns, tickets, and churn.
- Location pages: AI “service area” pages that repeat the same template with swapped city names are classic thin-content patterns.
- Programmatic SEO: large page sets that exist to rank but have no differentiated data or experience can trip thin-value signals.
A practical risk model: when AI content becomes “thin” in Google’s eyes
Let’s build a practical model you can use without being an SEO expert. AI content tends to become “thin” when these conditions stack together:
1) The scale is high
One AI-written article isn’t the issue. Ten thousand AI-written pages is a system issue. Scale amplifies every weakness: duplicated phrasing, missing nuance, incorrect generalizations, and thin internal linking.
2) The content is interchangeable
If you can swap your brand name with a competitor’s and the page still reads the same, you don’t have differentiation. Interchangeable pages are the easiest to classify as “little or no added value.”
3) There’s no evidence of lived experience
Especially for YMYL-adjacent topics (health, finance) and “hands-on” topics (repair, software configuration), content that reads like received knowledge with no proof of practice carries risk.
4) There’s no editorial accountability
Anonymous AI outputs with no author, no review, no editorial policy, and no update process create a trust gap. Even if “trust” is hard to quantify, the absence of process shows up indirectly in quality issues.
5) The site’s index footprint balloons
Index bloat is one of the most common “silent killers” in modern SEO: too many low-value pages dilute crawl attention and flatten sitewide quality perception.
Put differently: AI isn’t the crime. Unsupervised AI at scale is the risk.
Why this is happening now: AI search is forcing a higher bar
We’re entering a search environment where Google and other platforms want to answer the user directly, synthesize content, and cite sources. This changes incentives:
- In classic SEO, “good enough” pages could still win with links and technical health.
- In AI Search (AEO/GEO), content needs to be quotable: clear claims, specific steps, unique data, and credible authorship.
When AI systems summarize the web, they tend to ignore content that’s redundant. If your pages are AI-generated summaries of other summaries, they become citation-poor.
This is the new reality: Search visibility is increasingly about being a source, not a remix.
That’s why “thin content” and “AI generated” are converging in conversations. Not because AI is inherently low-quality—but because it’s commonly deployed to produce content that never becomes a primary source.
At AYSA, we talk to teams every week that are building for two outcomes simultaneously:
- Rank in traditional results (where relevancy and authority still matter)
- Get cited/mentioned in AI answers (where differentiation and structure matter)
Both goals demand that you stop thinking in terms of publishing volume and start thinking in terms of value density.
A concrete SME scenario: the local clinic with “helpful” AI pages
Here’s a realistic scenario I’ve seen variations of many times (details generalized).
A multi-location clinic wants to grow organic leads. They ask a marketer or agency to “create content for every service + every city.” AI makes this easy:
- “Sports physicals in City A”
- “Sports physicals in City B”
- “Sports physicals in City C”
Each page looks fine—800 words, FAQs, a call to action. The problem: the pages say essentially the same thing. They don’t include:
- Real clinician notes about who qualifies and who doesn’t
- Local insurance realities or appointment constraints
- What to bring, what forms to expect, what turnaround times are realistic
- What makes this clinic different (equipment, specialties, protocols)
Now scale that to 30 services × 25 cities = 750 pages. If the team also adds hundreds of AI-generated blog posts “to support the pages,” index size explodes. The site becomes 80% interchangeable content, 20% real business substance.
What happens next is predictable:
- Google crawls more, but indexes less reliably.
- Rankings become unstable.
- Leads plateau because the content doesn’t convert (it’s generic).
- And if a Manual Review happens—maybe via a spam report or a quality sweep—large sections can be deemed “little or no added value.”
The fix isn’t to delete everything. The fix is to consolidate, differentiate, and enforce quality gates.
What agencies must rethink (before clients get hit)
If you run an agency, this is a business model issue—not just a tactics issue.
Stop selling “content volume” as a deliverable
Clients used to accept “we published 30 posts” as progress. In 2026, that’s dangerously close to selling risk. The deliverable needs to become outcomes and signals: index quality, citation readiness, conversion lift, and content consolidation.
Build pruning and consolidation into retainers
Most retainers are built around production, not deletion. But “delete/merge/noindex” is often the highest ROI work you can do when AI has inflated the site.
Establish AI editorial policies clients can approve
Many teams are using AI without a clear policy. The policy needs to cover:
- What content types can be AI-assisted
- Where human experience is mandatory
- Disclosure standards (especially for UGC/community)
- Review workflow and accountability
- Testing and rollback procedures
Make “approval” a feature, not friction
AI makes it easy to ship. That means mistakes ship faster too. In high-stakes environments, the best agencies will differentiate on controlled execution.
This is exactly why AYSA’s model matters: propose changes, prepare them, and execute only after approval—at scale.
The 2026 thin-content prevention playbook (step-by-step)
This playbook is designed for SMEs and agencies who need practical steps—not theory.
Step 1: Inventory your index footprint (what exists, what matters)
You can’t fix what you can’t see. Start by mapping:
- Which directories/subfolders represent major content sets (e.g.,
/blog/,/threads/,/locations/) - Which sets drive revenue vs. vanity traffic
- Which sets were created via automation or AI
AYSA fit: Use AYSA Monitoring to track changes in indexable URLs, identify sudden growth patterns, and prioritize segments that look inflated or low-value.
Step 2: Classify pages by “added value” signals
Pick a sample set from each content segment and score it (simple, human scoring works):
- Does it contain original photos/screenshots?
- Does it cite primary or authoritative references where relevant?
- Does it include real steps, real constraints, and real outcomes?
- Does it demonstrate expertise (not just definitions)?
- Would a user bookmark it or share it?
If most pages fail, you likely have a systemic “thin value” problem.
Step 3: For forums/UGC, separate “community value” from “index value”
Not every thread should rank. That’s okay. Your community can still keep it for members, but Google doesn’t need to index it.
Practical options (choose based on your platform and business):
- Noindex low-value threads (duplicates, empty questions, unresolved one-liners)
- Merge duplicates into canonical “best thread” resources
- Close and summarize long threads with an editor-written “best answer” that adds value
- Elevate expert contributions and de-emphasize generic replies
- Re-think AI replies: if you use AI, clearly label it, limit it, and require human review for indexable answers
One of the hardest truths for forum operators: an AI reply that is “technically correct” may still be low value because it is not experiential. If the platform’s promise is authentic help, received knowledge isn’t a substitute.
Step 4: Create an AI publishing policy that forces differentiation
Here’s a policy template you can adapt:
- AI is allowed for: outlines, drafts, summarizing internal notes, extracting FAQs from support tickets, rewriting for clarity.
- AI is not allowed to publish without review for: medical/legal/financial advice, product compatibility claims, safety instructions, pricing promises.
- Every AI-assisted page must include at least 2 “added value” elements: original images, original data, expert quotes, real step-by-step, local constraints, test results, first-hand notes.
- Every page must have an owner: author/editor responsible for updates.
AYSA fit: AYSA can prepare changes and fixes while keeping humans in control—see AYSA AI SEO Tools and how we support execution without blind autopublishing.
Step 5: Structure content for both ranking and citation (AEO/GEO)
If you want visibility in AI-driven experiences, make your content easy to cite:
- Clear headings that match real user questions
- Short “direct answer” sections followed by nuance
- Step-by-step instructions with constraints and edge cases
- Explicit assumptions (“This applies if…”) and exclusions (“Don’t do this if…”)
- Where possible, references to primary documentation (vendor docs, standards, official guidance)
AYSA fit: This is part of our work on AI search visibility: content that isn’t just indexable, but “source-worthy.”
Step 6: Prune, consolidate, and “noindex” deliberately (quality > quantity)
The SEJ story triggered outrage about deleting threads without consent. In practice, you don’t need to delete content to keep it out of the index. Many sites:
- Keep content accessible to users while noindexing low-value pages
- Archive content behind navigation that doesn’t push it as a primary entry point
- Consolidate dozens of near-duplicate pages into one authoritative page
The goal isn’t censorship. The goal is to ensure that what Google indexes represents your best work and your real expertise.
Step 7: Fix internal linking so your best pages get the authority
Thin-value sites often have messy internal linking: thousands of pages linking to each other randomly. Consolidation only works if you funnel internal links to the canonical resources you want ranking and cited.
AYSA fit: AYSA can identify internal link gaps and propose structured fixes; you approve; AYSA implements at scale. That “approve then execute” workflow is the difference between controlled improvement and accidental sitewide chaos.
Step 8: Monitor quality signals continuously—not just rankings
Rankings are lagging indicators. The earlier signals often look like:
- Rapid growth in indexed URLs with flat impressions
- More pages discovered/crawled but fewer pages performing
- CTR dropping because snippets look generic
- Traffic shifting away from long-tail because content is redundant
AYSA fit: Use Monitoring to catch pattern changes early, then route fixes into an approval queue for execution.
If you got a thin-content manual action: triage and recovery plan
If you’re facing a manual action (or you fear you’re next), here’s a disciplined approach.
1) Confirm scope and isolate affected sections
Manual actions can be partial. Determine which folders, patterns, templates, or content types are implicated.
2) Identify the “dominant footprint”
In the SEJ example, the footprint was a forum section with hundreds of thousands of posts. In other businesses, it’s often:
- Location-page templates
- Programmatic long-tail pages
- AI-generated FAQ pages
- Auto-generated “support” pages
3) Apply fixes that remove the low-value pattern, not just individual pages
Google reviewers aren’t looking for a few improved pages; they’re looking for evidence you corrected the system that produced thin content.
System-level fixes include:
- Turning off autopublish
- Adding human review requirements
- Noindexing or consolidating large sets
- Adding expert-written summaries to community threads
- Improving author/editor information and page purpose clarity
4) Document what you changed
When you request reconsideration, you want a clear record of:
- What caused the issue (your best honest assessment)
- What you changed (with examples)
- What controls prevent recurrence
5) Expect iteration
Recovery can take time. The important thing is to move from “content production” to “content governance.”
The AYSA approach: approved execution beats “publish and pray”
Most teams don’t lose to “bad SEO.” They lose to execution debt:
- Too many pages to review
- Too many stakeholders to coordinate
- Too many templated patterns deployed too quickly
- No workflow for pruning and consolidation
AYSA is built for this reality. We operate like an execution system for modern SEO/AEO/GEO:
- Monitor site changes and risk signals: AYSA Monitoring
- Prepare specific fixes (content updates, internal links, indexing directives, consolidation plans)
- Ask for approval so humans stay accountable
- Execute accepted changes safely and consistently
That “approved execution” model is how you avoid the most common AI-content failure mode: a well-intentioned bulk publish that creates a low-value footprint you can’t unwind.
If you’re evaluating how to operationalize this, start here:
What to do next
Use this as your immediate action list for the next 14 days:
- Pick your biggest content segments (blog, location pages, help center, forum/UGC) and estimate how much of each was AI-assisted.
- Sample 20 URLs per segment and score them for “added value.” If you can’t find value quickly, Google probably can’t either.
- Identify your “interchangeable templates” (city swap pages, generic FAQs, auto replies) and pause autopublishing.
- Create a pruning plan: merge duplicates, noindex low-value, and elevate a smaller set of canonical resources.
- Rewrite the top 10 pages that should be cited (money pages + best informational pages) with real expertise, constraints, and proof.
- Set a governance workflow so AI can assist, but humans approve and own outcomes.
- Implement monitoring for index growth and performance drift so you catch “thin footprint” trends early.
Sources and further reading
- Search Engine Journal: Google May Be Penalizing AI-Generated Content As Thin Content
- Search Engine Journal: SEO section (ongoing context and reporting)
- Search Engine Journal: Google Algorithm Updates history (background timeline)
Note: This editorial is based on publicly discussed observations and reported events. Where the exact cause of a manual action is not verifiable from primary documentation, I’ve framed it as analysis and risk management—not certainty.
Continue the AI search topic inside AYSA.
Use these pages to connect the article with AI SEO tools, AI visibility monitoring, AI Overviews and approved website execution.
Turn this topic into a website action plan.
Use these AYSA hubs to move from reading to technical fixes, AI visibility monitoring, research, glossary context and approval-first SEO execution.
