+Writing

Field Notes

Practitioner writing on local SEO, AI search, and agent workflows. Everything here comes from work I actually shipped, with the receipts to prove it.

October 1, 2026

I Put SEOquake on Google's AI Overviews for Five St. Louis Searches. 74 Percent of What the AI Cited Was Google Business Profiles.

Semrush rebuilt SEOquake, its free Chrome extension, and on my install it put an Analyze sources button on top of Google's AI Overview. I ran it on five St. Louis home-service searches and exported every citation. 88 sources. 65 were Google Business Profiles. 19 were business websites, and when the overview was a written answer instead of a list, the pages it cited were service pages, not homepages. One of them belongs to a site with 11 backlinks.

Read the post ->

October 1, 2026

Your AI Visibility Score Is a Domain. The AI Cites a Page. I Counted Which Ones.

3,848 AI answers about 67 local businesses across ChatGPT, Perplexity, Gemini, and Google AI Mode. 786 times the answer cited the business's own site. Those citations landed on 206 pages, and the median business got 63 percent of its citations from one page. Across 16 sites with a full crawl, 85 of 820 pages were ever cited. Blog posts: one citation. Total. A domain-level number tells you none of this.

Read the post ->

October 1, 2026

I Sent Five AI Crawlers to 29 St. Louis Websites. Seven Slammed the Door.

I expected JavaScript to be the villain. It wasn't. I fetched 29 local business homepages as Chrome, GPTBot, OAI-SearchBot, PerplexityBot, and ClaudeBot, then rendered every one with Firecrawl. 22 let everybody in. Seven had a wall for at least one AI crawler: three Italian restaurants behind a Cloudflare challenge, a dentist that turns away ChatGPT's search bot, a law firm that throttles two of them, and a contractor with an expired certificate. Zero of them asked for any of it in robots.txt.

Read the post ->

October 1, 2026

I Wrapped Firecrawl in Nine Apify Actors. The Runs Lied. The Users Didn't.

Nine public Firecrawl actors, 6,307 runs, pulled from Apify's API. Then I looked at who was running them. The AI agent got the most users and the fewest repeat customers. The plain website crawler got the most runs. And one actor ran yesterday with zero users in 90 days. Here is what people actually want from a web scraper, and how my own app uses Firecrawl 45 files deep.

Read the post ->

October 1, 2026

Stop Publishing. More Than Half My Clients' Impressions Were Already Sitting in Positions 4 to 20.

I pulled the top 500 Search Console queries for five local clients, 2,500 in all, and sorted them by position. Queries in positions 4 to 20 held 54 percent of the impressions and earned 34 percent of the clicks. The top three held 9 percent of the impressions and took 64 percent of the clicks. The next win was never a new page. It was the pages already on page one and two.

Read the post ->

October 1, 2026

I Asked ChatGPT the Same 241 Local Questions Before and After the August Swap. Reddit Went to Zero. The BBB Took Its Seat.

Same 241 prompts, 62 local businesses, three monthly runs. After ChatGPT's August model swap, Reddit citations went from 35 to zero, the BBB went from 10 to 42, and answers with no sources fell from 73 to 11. The businesses' own sites got cited more. And 49 of 241 prompts flipped on whether the business was named at all. One snapshot tells you nothing. A fixed prompt set over time tells you almost everything.

Read the post ->

October 1, 2026

Page One Doubled. Clicks Didn't. I Sorted 4,944 Search Console Queries by Intent to Find Out Why.

Five local businesses, two 28-day windows, every query Search Console would give me, sorted into branded, local service, generic service, and informational. Page-one impressions on local service searches went up 129 percent. Page-one clicks went up 7 percent. Informational queries were 1 percent of the impressions to begin with. Half the clicks came with no query attached at all. The fight is not where the AI Mode headlines say it is.

Read the post ->

October 1, 2026

Google's Spam Update Is Still Rolling Out. I Checked 68 Local Rankings Before Blaming It.

Google started the September 2026 spam update on September 24 and gave it up to two weeks. Before anyone blamed it for anything, I lined up 68 local map rankings scanned on both sides of the start date and five Search Console accounts. The map barely moved more than it moves in a normal month, the organic trend did not bend, and the biggest drop I found has a better suspect than Google.

Read the post ->

September 25, 2026

A 95-Step Zap Won Me a Trophy. My Newest Zapier Workflow Has Zero Steps.

Project Phoenix, a 95-step Zap, earned Zapier's 2025 Automation Hero of the Year. This month I built a Zapier workflow with no steps at all: an agent on a schedule, a saved Zapier MCP skill, and four connections that brief me every weekday morning. Both are right. Here is how to tell which one you need.

Read the post ->

September 25, 2026

It Took Me 55 Versions to Learn One Make Lesson: Stop Waiting

God Mode Intel made the finals of Make's MCP community challenge. The folder behind it has 55 private scenario versions built in three days, with names like minimal, barebones, and ultra-simple. The one that worked had two scenarios and a callback, and nothing in it waited for anything.

Read the post ->

September 25, 2026

I Published 283 Apify Actors. People Mostly Run the Boring Ones.

283 public actors, 8,584 users, 183,095 runs, pulled from Apify's public store API. The flashy actors get tried once. The dull ones get scheduled. My most-run actor detects what software a website uses, and the runner-up scrapes the Better Business Bureau.

Read the post ->

September 22, 2026

Ask an AI What Your Business Is and It Reads You. Ask Whether to Trust You and It Reads Everyone Else.

Six questions about three of my clients, three engines, 210 sources. When the question was what the company is, 47 percent of the sources were the company's own site. When the question was whether to trust it, 80 percent were somebody else: Google's review pane, Yelp, the BBB, a Birdeye page nobody remembers setting up. Ranking gets you the introduction. Third parties do the vouching.

Read the post ->

September 22, 2026

I Ran an AI Detector on Ten Pages That Rank. It Flagged Six. It Flagged Me Too.

Detectors are coin flips, the watermarks are moving inside the models, and Google now rejects review replies for sounding scripted. I pointed a structural tell detector at ten top-ranking pages and four of my own. It flagged the humans and it flagged the human. The only durable defense is effort you can show, and the rejection data says exactly what that looks like.

Read the post ->

September 22, 2026

Nobody Copies Anymore. Everybody Agrees. I Measured It.

I pulled the top results for two questions my clients get asked every week and measured how alike the pages are. No copied sentences. The same outline, the same seven dollar figures, the same order. The AI Overview cited the page with the most specific numbers and skipped the two biggest brands on the page. Original is not a vibe. It is a number you can add.

Read the post ->

September 22, 2026

Your Keyword Map Needs a Prompt Column. I Built One for Three Clients.

I rewrote real client keywords as the questions people actually ask, tagged each by funnel stage, and ran them through ChatGPT and Perplexity. Every who or best prompt produced a list of businesses. The how much and should I prompts produced numbers and advice and, nine times out of ten, no business at all. That is the whole planning model, and it fits in one extra column.

Read the post ->

September 22, 2026

I Tested Whether AI Cites Boring SEO. It Does. Mostly.

Seven real client keywords, four AI engines, one afternoon. The engine that reads the open web cited the client every time. The engines that read Google's own surfaces cited the map pack and the Better Business Bureau. The work that earns a citation turns out to be the work you were already supposed to be doing.

Read the post ->

September 22, 2026

Google Hides Half Your Clicks. I Counted.

Five client sites, twenty-eight days, one Search Console pull. The page report explained every click. The query report explained 52 percent of them. Google calls the rest anonymized, which is a word it uses so you will not use the other one.

Read the post ->

September 10, 2026

"Crawled, Currently Not Indexed" Is Google's Polite Way of Saying No

Google now filters for quality before your page ever gets a chance to rank, and it says so on the record. Here is what that means for the fifty AI-written city pages a previous agency shipped, why Search Console lies to you about it, and what a location page has to earn to get in.

Read the post ->

September 9, 2026

Your Plumber Is a YouTuber Now. The AI Made Him Do It.

AI answer engines cite YouTube more than any other social platform, and Google puts videos on the money queries where contractors used to live. Here is what that means for a business with a truck, what the search results actually show, and the forty-second video that beats your blog.

Read the post ->

September 8, 2026

What Do I Do When AI Gets My Brand Wrong?

Ask an assistant about your business and it hands back a blurry average of every name, address, and claim the web holds about you. Here is the six-step fix I ran across a whole client fleet last week, the part that was our own fault, and where Semrush fits.

Read the post ->

September 8, 2026

How Do I Do Keyword Research for Multiple Locations?

Not by finding-and-replacing the city name. Five suburbs, five different people typing five different things. Here is the per-location process I ran for a St. Louis brokerage last week, the Semrush steps, and the rollup that keeps a franchise owner and a location manager looking at the same numbers.

Read the post ->

September 8, 2026

How Do I Build Local Citations at Scale?

Slowly, correctly, and with less enthusiasm than the word scale implies. A citation is your name, address, and phone on a site you do not control, and the failure mode is a client listed at the wrong suite for years. Here is the audit-first process I run, the scraper fleet behind it, and where Semrush fits.

Read the post ->

September 8, 2026

Your Customer Asked One Question. Google Asked Ten More. You Answered One.

AI search does not look up your keyword. It breaks the question into a pile of smaller ones you never ranked for, then cites whoever answered those. Here is what fan-out does to a local service business, with the real sub-questions pulled from Semrush, and how to stop losing a job you were never in the running for.

Read the post ->

September 2, 2026

Your Analytics Dashboard Is Lying to You About Where Customers Come From

AI assistants are sending you customers and taking credit for none of it. Your traffic looks flat, your phone keeps ringing, and your dashboard has no idea why. Here are the KPIs that actually track AI search, and the ones quietly lying to your face.

Read the post ->

September 1, 2026

I Gave an AI Agent One Job: Find My Competitors. Here Is the Raw Output.

Everyone demos MCP with a toy question. I ran one real use case through the Semrush MCP against a live vinyl wrap shop and pasted the unedited results. Fifteen competitors, ranked, in one tool call. Here is what it found and where it still needs a human.

Read the post ->

September 1, 2026

ChatGPT Stopped Searching the Web. Does It Already Know Your Business?

In one model update, ChatGPT quit searching the open web and started picking a shortlist of trusted domains first. Getting found is no longer the game. Getting named is. Here is what that means for a local business.

Read the post ->

August 29, 2026

Your Website Makes Claims. Entity Corroboration Decides If They Are True.

Page optimization is losing to entity optimization. AI systems no longer take your site's word for anything, they cross-check it against the rest of the web. For a local business, that is a gift, if you know what to fix.

Read the post ->

August 27, 2026

MCP Agent Workflows Are Replacing Manual SEO. I Run Them in Production.

Most takes on agentic SEO are speculation. This one comes from an agency that wired the Semrush MCP into its daily client work. Here is what actually changed, and where the humans still earn their keep.

Read the post ->

August 27, 2026

AI Overviews Are Ending the Click. Local Businesses Can Still Win.

Zero-click search sounds like doom until you notice who actually loses. A field guide for local and service businesses on staying visible when the answer arrives before the website does.

Read the post ->