How many pages should a small business have indexed on Google?

A small business should have as many pages indexed as it has distinct, high-quality topics to cover, which for most is between 30 and 150. Quantity isn't the goal. A site with 50 strong pages almost always outperforms a site with 500 weak ones in 2026.

Does Google detect AI-written content directly?

Google doesn't directly penalize AI-written content based on detection. It penalizes content that lacks originality, expertise, or first-hand experience, which AI content tends to lack by default. A skilled human using AI as a tool can absolutely rank. A prompt-and-publish workflow typically can't.

What counts as thin content in 2026?

Thin content in 2026 is any page that doesn't add unique information, perspective, or experience beyond what already exists online. A 2,000-word post can be thin if all 2,000 words are generic summary. A 400-word post can be excellent if those 400 words contain something no one else has said.

Should I delete old blog posts or just update them?

Delete posts that have no traffic, no business purpose, and no realistic path to either. Update posts that have a real topic and a real audience but outdated information. The mistake most businesses make is trying to update everything, which spreads effort too thin to matter.

How long does content recovery take after a Google update?

Content recovery after a Google update usually takes 30 to 120 days once you've made the changes, and sometimes longer if the site was hit hard. Google doesn't re-evaluate sites in real time. Consistency of improvement signals over weeks is what triggers recovery.

Is it better to hire a writer or use AI with human editing?

For most small and mid-sized businesses, a skilled human writer who uses AI as a research and drafting tool produces the best results per dollar. Pure AI is too generic to rank. Pure human takes too long to scale. The hybrid model wins when the human brings real subject matter expertise to the work.

The Content Volume Paradox: Why More AI Content Kills Rankings

A friend of mine runs a regional services company. Last year he got talked into publishing 400 AI-generated blog posts in six months. His organic traffic didn't go up. It went down 62 percent. He called me panicking, thinking he'd been hit by a manual penalty. He hadn't. Google had just done exactly what Google said it would do. This is the content volume paradox, and it's punishing more businesses every month. I'm Jake McCluskey, and after 25 years in digital marketing and working with 500+ businesses, I can tell you the advice to "publish more AI content" is the single worst piece of SEO advice circulating right now.

Why is publishing more AI content hurting rankings instead of helping?

Publishing more AI content hurts rankings because Google's systems now treat low-effort, undifferentiated pages as a signal of site-wide quality problems, not just individual page problems. When you flood a domain with thin AI output, the whole site gets devalued, including your money pages. It's the opposite of what the 2019 playbook said to do.

Here's what changed. Google's helpful content system, rolled into the core ranking systems in March 2024, evaluates quality at the site level. If 70 percent of your URLs look like recycled summaries of things other people already said, Google treats the whole domain as a low-value source. Your one good page about your actual service gets dragged down with the slop.

I've watched this happen to four clients who ignored the warning. One of them had 1,200 blog posts and 14 service pages. After we deleted 900 of the blog posts, rankings on the service pages recovered inside 90 days. The traffic math got better by subtracting.

The paradox is that volume used to be a competitive moat. It isn't anymore. Volume is now a liability unless every piece earns its place.

What do Google's helpful content and E-E-A-T signals actually penalize?

Google's helpful content and E-E-A-T signals penalize content that reads like it was written by someone who doesn't know the topic, doesn't use the product, and didn't do the work. The algorithm isn't testing for "AI or human." It's testing for experience, expertise, and originality. AI content fails those tests by default.

E-E-A-T stands for Experience, Expertise, Authoritativeness, and Trust. Google added the second E (Experience) in December 2022 specifically because generative AI made the first three easier to fake. Experience is the hardest signal to spoof. It shows up in the details: the screenshot of the thing you actually did, the specific price you actually paid, the mistake you actually made.

AI content written from a prompt like "write 1,500 words about CRM software" has none of that. It's summary. It's vibes. It's the average of what everyone else wrote, which is exactly what the algorithm is trying to filter out.

Specific things that get penalized:

Pages that answer questions without unique data, examples, or screenshots
Pages with no clear author or an obviously fake author bio
Clusters of posts that all sound identical in tone and structure
Content that contradicts itself within the same site because different prompts produced different takes
Pages that don't reflect any first-hand use of the product or service discussed

If your content would be equally useful coming from any random website in your industry, it's the kind of page Google is actively trying to demote.

How does LLM-era search punish thin content even harder?

LLM-era search punishes thin content harder because answer engines like ChatGPT, Perplexity, and Google's AI Overviews cite sources based on specificity and originality, not just traditional SEO signals. If your page says the same thing as 40 other pages, you're invisible. The LLM picks the one source with a unique angle or a unique data point.

I ran an experiment in February. I took 20 client pages and fed their core queries into ChatGPT, Perplexity, and Google AI Overviews. The pages that got cited had one thing in common: a specific number, a named example, or a first-person claim. The ones that didn't get cited were all well-optimized by 2019 standards. Clean H1s. Proper schema. Good internal linking. None of it mattered.

LLMs are doing extractive summarization. They need something to extract. If your 1,800-word post is 1,800 words of general-interest padding, there's nothing to pull out.

This changes the economics completely. In the old world, a decent post might catch some long-tail traffic even if it wasn't great. In the AI-answer world, a decent post gets zero visibility because the LLM quotes the best page, not the tenth-best. The gap between first and tenth used to be maybe 10x in traffic. Now it's closer to infinity to zero.

What does the right publishing cadence look like in 2026?

The right publishing cadence in 2026 is fewer, better, deeper pieces, published on a schedule you can actually sustain with real expertise behind each one. For most small and mid-sized businesses, that's 2 to 4 posts per month, not 20. Volume is no longer a strategy. It's a failure mode.

Here's what I recommend based on business size and resources:

Solo operator or very small business (under 10 employees): 2 posts per month. Each one 1,500 to 2,500 words, based on something you actually did, saw, or tested. One pillar article per quarter at 3,000+ words.
Small business (10 to 50 employees): 4 posts per month. At least two should include original data, case studies, or expert interviews. One major piece per quarter.
Mid-sized business (50 to 250 employees): 6 to 8 posts per month maximum. You have enough subject-matter experts to sustain this if you use AI to assist, not to replace. One big research piece per quarter.
Anyone publishing more than 12 posts per month without a real newsroom-style team: you're probably hurting yourself.

The key shift is treating each post as a real investment rather than a checkbox. A post that takes 6 hours and includes original thinking will outperform 30 AI-generated posts that took 6 minutes each. I've measured this across dozens of sites.

If you're not sure what you currently have, start with a free audit and figure out which pages are pulling weight and which are dragging you down.

Is AI useless for content then? Should I just write everything by hand?

AI isn't useless for content. It's useless as a replacement for thought. Used correctly, AI speeds up the parts of content creation that don't require original expertise: outlining, first-draft scaffolding, grammar polish, variant generation. Used incorrectly, it produces the exact kind of page Google and LLMs are trained to devalue.

The pattern I've seen work with clients is a roughly 30/70 split. AI handles about 30 percent of the labor: research synthesis, outline drafts, headline options, editing passes. The human handles the 70 percent that actually determines whether the page ranks: the angle, the original data, the specific examples, the opinion, the voice.

A useful test: if I removed everything from the post that came from your personal experience or your business's data, is there anything left? If the answer is "not much," you've got an AI-slop problem even if a human technically pressed publish.

I use AI every day in my work. I'm not anti-AI. I'm anti-slop. There's a real difference.

How do I recover if I've already published a ton of thin AI content?

If you've already published a ton of thin AI content, you recover by ruthlessly pruning, consolidating strong pages, and slowing your publishing cadence. The fix isn't writing more. It's writing less and deleting most of what you already have. This feels wrong to most business owners, which is why most of them won't do it and will keep losing traffic.

The recovery process I've run for several clients:

Audit every indexed URL. Pull your top 1,000 URLs from Google Search Console and sort by impressions over the last 90 days.
Flag the bottom performers. Any page with under 10 impressions per month is a candidate for deletion or consolidation. Be honest about whether it has any purpose at all.
Merge overlapping topics. If you have six posts about "email marketing tips," combine them into one strong pillar page and 301 redirect the rest.
Delete the true junk. Not noindex. Delete. Return 410 codes so Google stops crawling them. Noindex keeps them in the crawl budget, which is still a drag.
Rewrite the pages that matter. Take your 10 to 20 actual money pages and rewrite them with real examples, real numbers, real voice.
Wait 30 to 90 days. Google needs time to re-evaluate the site. Don't panic-publish during this window.

Clients who run this process typically see recovery start around day 45 and plateau around day 120. It's not fast. It does work.

If pruning your own site feels like surgery without a mirror, that's what our services are built for. I'll tell you exactly what to cut.

Why aren't more marketers talking about this?

More marketers aren't talking about this because a lot of them sell AI content tools, AI content services, or "publish 100 blog posts a month" packages. The incentive to push volume is enormous because volume is easy to sell. Quality is harder to sell because it sounds slower and more expensive, even though it produces better results.

I don't sell a content volume package. I don't sell AI content subscriptions. When I tell a client to publish less, I'm not upselling them on anything. That's the whole point. The advice you get from someone who gets paid more when you publish more is not the advice that's going to help you rank in 2026.

Read the helpful content documentation from Google directly. Read the Search Quality Rater Guidelines. Both are public. Both say, in plain language, the things I'm saying in this post. The only people pretending otherwise are the ones selling you the problem.

There's also a simpler tell. Ask the agency or tool vendor pushing high-volume AI content to show you their own rankings. Look at the traffic for the sites they publish on themselves. In most cases, you'll find domains running the same strategy they sell are also losing traffic. They just don't lead with that slide.

The businesses winning at content right now publish less, say more, and stake their reputation on every piece. They treat their blog like a portfolio, not a factory. A portfolio has 20 strong entries, not 2,000 mediocre ones. That's the mental shift that fixes the problem. If you're still measuring content output in posts per week instead of pages that actually earn traffic and leads, you're measuring the wrong thing entirely. I've watched clients cut their publishing schedule by 80 percent and double their organic traffic inside six months. It's not magic, it's just taking the signals Google keeps telling us about seriously. If you want a second set of eyes on what's working and what's quietly costing you, book a discovery call and we'll look at your site together. I'll be straight with you about what to keep, what to cut, and what to rewrite, and I won't sell you more content you don't need.