Brass-SEO

Semrush costs $139.95/month. For most small businesses, that's a lot to justify. I've been digging into whether it actually delivers ROI for smaller teams — and the honest answer is: it depends on what you're trying to do. Semrush is one of the most recognized names in SEO software. It's also one of the most expensive. If you're trying to figure out whether it's worth the cost for your business, I've broken down what each plan includes, who it's actually built for, and where the real value sits. Here's what I found: **The Professional Plan ($139.95/month)** covers keyword research, site audits, and basic rank tracking. Solid for a single person or small agency running their own SEO. But the cap matters — limited projects, limited integrations. **The Business Plan ($249.95/month)** is where you get breathing room. More projects, more users, better API access. This is where solo operators typically upgrade when they take on clients. **The Enterprise tier** starts at $499+/month and targets mid-size agencies. White-label options, dedicated support, custom workflows. The gap between tiers is steep. You're not just paying for more features — you're paying for how many projects, how many users, and how much data you can actually work with. A freelancer and a 10-person agency need different things. But here's the question most small business owners don't ask: do you need Semrush, or do you need *some* SEO tool? There are cheaper alternatives. Some are weaker. Some solve the same problems differently. The post walks through what each tier actually gives you, the hidden constraints, and honest advice on whether the entry plan makes sense for your revenue stage. Read the full post → https://lnkd.in/gZ_3y5Yd

How do you audit technical SEO without dropping $500/month on a tool? Most AI prompts for technical SEO are built for ChatGPT or generic LLMs. But Claude has one superpower that changes what's possible: a 200,000-token context window. That means you can paste your entire crawl log, full robots.txt, complete sitemap, AND multiple pages of HTML into a single conversation — and get analysis that sees all of it at once. Most technical SEO tools require you to buy a subscription to do what Claude can do in one conversation. The trick is knowing what to ask Claude and how to frame it so you get actionable answers, not just generic observations. A good prompt does three things: (1) it tells Claude exactly what it's looking at, (2) it specifies the output format you need, and (3) it surfaces the problems that matter — crawlability, indexability, schema, log file patterns — without overwhelming you with noise. We've built 12 copy-paste prompts that cover the full technical SEO audit spectrum. Each one is designed specifically for Claude's strengths — its ability to hold massive context and reason about patterns across thousands of lines of data. They cover: - Crawlability analysis (what Googlebot can and can't reach) - robots.txt and crawl directive review - Sitemap XML validation - Log file analysis (yes, you can feed Claude server logs and it spots patterns in seconds) - Schema markup audits - Redirect chain detection - Mobile usability signals - And more. Each prompt includes the exact wording, what to paste into Claude, and what to look for in the output. No guessing. If you've been using generic AI prompts and getting vague answers, or if you've been thinking "I should audit this but the tool subscriptions are too expensive," this is worth 10 minutes of your time. Read the full post → https://lnkd.in/gMYk-3ip

We published a guide on the Grokipedia correction strategy — find a factual error, publish authoritative content, submit a correction, earn a backlink. Sounds straightforward, right? Then we actually ran the playbook ourselves. We searched 12 Grokipedia pages looking for correctable claims. The question: how many real opportunities exist, and what separates a genuine correction from a dead end? Turns out, there's a gap between "this claim is wrong" and "this qualifies for a correction submission." We found plenty of outdated statements and dubious facts. But most didn't meet the bar. Some lacked authoritative sources to point to. Others were opinions, not factual errors. A few were technically wrong but so niche that publishing content around them made no sense for SEO. Only one submission made the cut — the one we actually filed. This report breaks down exactly how we filtered 12 pages, what we looked for, what disqualified candidates, and why that single correction was worth the effort. If you've been wondering whether Grokipedia is a viable link-building channel for small businesses (spoiler: it can be, but not how you might think), this shows you the real math. Read the full post → https://lnkd.in/g6T4dNSb

What if you could earn backlinks by helping fix mistakes on one of the internet's fastest-growing reference sites? Grokipedia—X's AI-generated encyclopedia—produces reference-style articles on thousands of topics. And like most AI content, it has inaccuracies. But here's the opening: when users flag incorrect information and cite a source, Grokipedia's review process can update the article and link to that source. That means a backlink to your site. Most people don't know this opportunity exists. Even fewer have built a systematic approach to finding and correcting errors at scale. Our sister project, BrassTranscripts, tested this in November 2025. They identified inaccuracies in Grokipedia articles related to their niche, submitted corrections with their source cited, and watched Grokipedia accept the updates. Each acceptance came with a backlink from an X-hosted property—a link most SEOs overlook entirely. The strategy works because it solves a real problem for Grokipedia: AI-generated content needs human review to stay credible. You're not gaming the system; you're contributing value to a platform that's actively looking for corrections. The hard part isn't the concept. It's finding the right errors in the right articles—the ones where Grokipedia will actually accept your correction and where your site's authority matters for the link. You need to know which topics connect to your niche, which articles actually contain errors, and how to frame a correction in a way the platform approves. BrassTranscripts walked through the entire playbook: how they identified opportunities, what their correction process looked like, which submissions were accepted (and which weren't), and what they learned about timing and topic selection. If you're building backlinks through resource creation or guest content, this is a complementary angle most competitors aren't touching yet. Read the full post → https://lnkd.in/gcZsa-vr

No published research has directly measured whether schema markup increases AI citations. But here's what we do know: AI systems like ChatGPT, Perplexity, and Google AI Overviews draw from web content in ways that schema *may* help — the evidence is just indirect. I've been digging through what's actually been studied. The gap is real. We don't have a number on schema's impact for AI search. Not yet. What we *do* have is a clear picture of what schema does, how AI systems read web content, and where the two intersect. That matters because the decision to add schema markup — or skip it — depends on understanding that intersection. If you're deciding whether to implement schema for AI search reasons, the honest answer is: it cannot hurt, it probably helps, and we do not have a definitive impact number. But the research that *does* exist points to how schema could matter — and why the absence of direct measurement doesn't mean absence of value. Read the full post → https://lnkd.in/gSmkm3Kr

Most SEO content fights on identical ground. The same Semrush exports, the same Google studies, the same industry reports — all rewritten in slightly different order. Every "best practices for X" post competes with every other "best practices for X" post using the same source material. Original data posts break that pattern. They contain findings no competitor has, because the underlying records belong to you. We recently published two data-driven posts using six months of real production database records — no surveys, no extrapolation, no estimates. Just actual behavior from a live system. Here's what made it work: **The methodology.** We pulled directly from production logs and real user interactions. No guessing, no modeling. The data was messy, unfiltered, and authentic — exactly what search engines reward. **What the data showed.** The posts revealed patterns competitors can't access. Traffic movements, user behavior shifts, content performance across seasons. Insights that only emerge when you're working with your own records, not borrowed case studies. **Why it's unreplicable.** Your competitors see the published posts. They can't see your database. They can't run the same analysis on their own production data without months of setup and the discipline to do it right. That asymmetry is an SEO advantage that compounds. The posts ranked because they answered questions no one else could answer with that level of specificity. Search engines noticed. Users noticed. If you're building SEO content, this is the moat to chase: original data from your own systems. Not surveys. Not borrowed studies. Your records. Your insights. Your competitive edge. Full breakdown → https://lnkd.in/gJF_Mt4N

How much AI traffic is actually hitting your site? If you're only checking GA4, you're seeing maybe half the picture. GA4 catches the referral clicks — when someone reads an AI-generated answer that links to your site and clicks through. You'll see `chatgpt.com / referral` and `perplexity.ai / referral` sessions showing up in your reports. That's real traffic, real humans arriving from AI tools. But GA4 misses the crawls. Server logs capture every HTTP request, including GPTBot, ClaudeBot, PerplexityBot, and dozens of other AI crawlers fetching your pages directly. These bots are spidering your content before any citation appears in an AI answer. GA4 will never show you these requests — they're not human clicks, so the analytics tool filters them out. Here's why both matter: The crawler activity tells you which AI models are indexing your content and how often. That's an early signal of future citations. The referral clicks tell you which AI tools are actually sending you traffic *right now*. One is leading indicator, one is lagging. Ignore either one and you're flying blind. If you only look at GA4, you'll think AI traffic is smaller than it really is. You'll miss the bots. If you only dig into server logs, you won't know whether all that crawling activity actually converts to citations or referral visits. You need both signals to understand the full AI-to-traffic pipeline on your site. The post walks through what each tool catches, which AI models are most active on your site, and how to spot the difference between a bot that's indexing your content and one that's actually sending you visitors. Read the full post → https://lnkd.in/gFRjstvQ

What makes AI systems cite your content instead of someone else's? Princeton's GEO research tested nine different content modifications and measured their effect on AI citation rates. Adding expert quotes produced a +41% lift — the single largest improvement of any strategy tested. No other tactic came close, not statistics, not improved fluency, nothing. But here's the gap: that finding gets repeated everywhere, almost always without the detail that actually matters. What counts as an "expert quote" in the Princeton study? Why does it move the needle so hard? Which types of quotes work, and which ones don't? (Spoiler: a quote from someone with a Twitter account isn't the same as one from someone who actually has skin in the game.) The difference between a quote that lifts your citations by 41% and one that does nothing is specific. It's about authority, it's about how you present it, and it's about which experts your target AI system recognizes. If you're writing content that needs to be cited by AI — whether that's for SEO, thought leadership, or just being the source people (and systems) trust — this matters more than the usual optimization moves. A well-placed expert quote can outweigh a hundred other tweaks. Read the full post → https://lnkd.in/ga5Dv5eu

Does longer content get cited more by AI systems? It's the question everyone assumes has an obvious answer. But here's what the research actually says: word count doesn't matter. Not directly. I've pulled the specific studies — the Gnuse 2025 analysis of 7,500 ChatGPT referral sessions, the Princeton GEO study by Aggarwal et al., and others — and none of them isolated word count as a variable affecting citation rates. The assumption that "longer is better" comes from extrapolating Google ranking research, not from testing how AI systems actually choose what to quote. What *does* correlate with citations is something else entirely. Something most creators are ignoring. The pattern researchers found isn't about bulk. It's about structure. It's about whether a piece of content is self-contained — whether it answers a complete question without forcing the reader to jump between sources. It's about how the information is organized so an AI can extract a coherent claim without hunting for context. This matters because it changes how you should think about your content strategy. The instinct to add 2,000 words "for SEO" — or to pad sections with filler — isn't supported by what we know about AI citations. But creating content that stands alone, with clear sections and substantive examples that a system can reference without ambiguity? That's where the signal is. The distinction is subtle but real. And it flips how a lot of creators approach their work. The full post walks through the studies, names the specific metric that *does* predict citations, and shows you what that looks like in practice — the structural patterns that appear in cited content versus what gets skipped. Read the full post → https://lnkd.in/gV8JuYvE

Google Search Console shows you exactly which pages are losing clicks — but most people ignore it. You've got pages ranking in positions 6–10, getting zero clicks. Pages that used to rank higher but slipped. Pages with high impressions and terrible click-through rates. They're right there in GSC, waiting to be fixed. Here's what most SEOs do: panic and write new content. Here's what actually works: spend 30 minutes refreshing what you've already got. Refreshing beats writing from scratch because your domain already has authority on the topic. Google already knows the page exists. You're not starting at zero — you're moving a page from position 8 to position 3. The workflow is simple: Step 1: Open GSC and filter for pages with high impressions but low CTR (or pages in positions 6–10). You're looking for traffic that's within reach. Step 2: Diagnose why. Pull the top 3 queries for that page. Are the results above you more recent? Do they have better formatting? Is your meta description vague? The SERP tells you what Google is rewarding. Step 3: Make one or two targeted fixes. Update the intro. Add a fresh statistic. Restructure with a better headline or subheads. You're not rewriting — you're sharpening. Step 4: Tell Google you refreshed it. Update the publish date in your CMS, hit refresh in GSC's URL inspection tool, and request indexing. Done. The entire process takes 30 minutes per page if you're disciplined. Pick your lowest-hanging fruit first — pages that are already close to the top but just need a nudge. The blog post walks through the exact GSC filters to use, how to read the SERP to diagnose problems, and what type of refresh actually moves the needle (spoiler: it's not always content length). Read the full post → https://lnkd.in/gVbmhdNK