About
👋 Hi, I’m Leigh.
I help brands grow visibility, trust, and revenue across Google,…
I help brands grow visibility, trust, and revenue across Google,…
Articles by Leigh
-
How to Find and Close AI Citation Gaps in 2026: A Step-by-Step Guide
How to Find and Close AI Citation Gaps in 2026: A Step-by-Step Guide
It’s not uncommon for competitors to show up in ChatGPT answers, Google AI Overviews, and Google AI Mode results…
10 Comments -
How to measure AI share of voice using Semrush
How to measure AI share of voice using Semrush
If you're wondering if AI platforms are mentioning your brand or a competitor's, you need to be looking at AI share of…
18 Comments -
Reddit Marketing in 3 Hours/Week: From 0 Karma to Real Cred
Reddit Marketing in 3 Hours/Week: From 0 Karma to Real Cred
This article was originally published on Backlinko. You can read it there.
8 Comments
Activity
37K followers
-
Leigh McKenzie shared thisGetting up to speed on AI search costs nothing. Here are 11 free resources, in the order I'd work through them: 1. Benchmark your AI visibility ( Semrush ). Enter a domain, see how often you show up in ChatGPT, Gemini, and Perplexity versus competitors. https://lnkd.in/eTbrmECs 2. Read the 2026 AI Visibility Index. 126 million prompts. The closest thing to a census of who gets cited and why. https://lnkd.in/eYpeuxMg 3. Set up Bing Webmaster Tools and Microsoft Clarity. Bing feeds Copilot and ChatGPT, and its AI performance report is first-party data most teams are ignoring. https://lnkd.in/eWb_ap_U 4. Get clear on what GEO actually is (and isn't). https://lnkd.in/ed2BBpD8 5. Sign up for the AI Search Operating System. A free course for people who own a process, not just a site. https://lnkd.in/eiubq63R 6. See where the agentic web is going. WebMCP is the part of this most SEOs haven't read about yet. https://lnkd.in/e5PaqXkm 7. Prepare for agentic commerce. If you sell anything online, protocols like UCP will decide whether agents can buy from you. https://lnkd.in/eDUmAXb5 8. Learn how query fan-out works. This is the mechanism behind most of what looks random in AI answers. https://lnkd.in/ewPaSyzy 9. Do prompt research for AI SEO. Keyword research still matters, but prompts are a different unit of demand. https://lnkd.in/eZgUgGec 10. Understand topical gravity. AI visibility is won at the topic level, not the page level, and this explains why. https://lnkd.in/e3aPdwXY 11. Keep up with the news. Search Engine Land covers every AI search development the day it happens, and it's where I check first when something changes. https://lnkd.in/eQRGMQgx Run number 1 before you read anything else. It takes a domain and a click, and it tells you whether the rest of this is urgent or optional.
-
Leigh McKenzie posted thisThere's a whole cottage industry now selling AI search like some brand new dark art. Fresh acronyms. New courses. Frameworks with names.... I get why, because a shiny new discipline sells a lot better than telling someone to go fix their HTML. What actually decides whether an AI system can pull your content into an answer is the same technical SEO that's been on the checklist for over a decade. 1- Can the crawler reach the page 2- Is the HTML clean enough to read 3- Does the schema tell the machine what it's looking at 4- Is the content fresh and organized like a human structured it on purpose What changed is the cost of getting it wrong. A stray noindex or a broken redirect chain used to cost you some rankings. Annoying, contained, fixable on a Tuesday. Now that same mistake can wipe you off Google and out of every AI surface pulling from the web, all at once. One technical mess, every front door shut. The part barely anyone is ready for is what happens when agents show up to buy. In-chat checkout stalled this year and the money moved back to your own site, which means the agent is coming through your storefront. You're out before it starts if: Your checkout falls apart the second something moves through it without running your JavaScript. Your product schema says in stock when it isn't. Your add-to-cart is a custom div a machine can't even see. Go try to buy something on your own site the way an assistant would send someone through it. Watch where it stalls. That stall is a lost sale. So before you buy the course on winning AI search, open your robots.txt and read what you're blocking. Look at your HTML and ask whether a machine could actually parse the thing you spent six months writing. The boring work is the work. Recommended article: https://lnkd.in/eEzvrdYS ("The technical SEO checklist for search engines and AI search" by Semrush)
-
Leigh McKenzie shared thisWhich off-page tactic still works in 2026? On-page and technical SEO only control what happens on your domain. Google and AI systems also read what the rest of the web says about you before deciding who to surface. How Id structure off-page work right now: 1. Earn links from relevant sites Relevance over volume. A few links from sites your buyers read do more than a pile from directories. → Guest posts on sites your audience reads → Outreach to writers covering your topic → Original data people have to cite 2. Track unlinked brand mentions AI models read the sentence around your brand name, so a mention with no link still counts. Where to check: → Industry publications and newsletters → Reddit, Inc. and niche communities → Podcasts and YouTube transcripts → AI answers to category-level questions 3. Study why competitors get cited Pull their referring domains and the sites that mention them without linking. Then prompt an AI assistant with your buyers questions and note who gets named. Every gap is a target. 4. Distribute where your audience already is Posting the piece on LinkedIn, X, and the two or three communities your buyers use is how it reaches the people who eventually link to it. Nobody links to a page they never saw. 5. Local signals still count For anyone with a physical footprint, keep your business details identical everywhere. → Same name, address, and phone everywhere → A complete Google Business Profile → Mentions on local sites and directories AI tools recommending a local business lean on these same sources. All five feed the same thing: a consistent record of your brand across the web. That record now decides whether you rank, and whether an AI answer bothers to name you at all. This visual concept was inspired by Ankush Gupta
-
Leigh McKenzie posted thisEveryone in SEO spent the last year trying to get cited. Almost nobody checked what else is sitting in the pool they're getting cited alongside. A new audit ran 712 real user questions on politics, health and the environment through ChatGPT, Microsoft Copilot, Gemini and Perplexity, then scraped every source those engines linked to. 26,266 unique URLs across 7,675 domains. The researchers pulled the text from the ones they could reach and ran all of it through an AI detector. Roughly 16% came back machine-written. The authors call that a floor rather than a ceiling, because they deliberately threw out the ambiguous cases and their detector misses about a third of confirmed synthetic text. The spread by engine is the part I want us to focus on. Copilot cited AI-generated sources in 27.8% of its citations. Gemini, 14.7%. Perplexity, 9.4%. ChatGPT, 7.3%. Same web, four very different filters, and nothing on screen tells a user which filter they're looking at. 97% of those synthetic sources sat outside the 25 most-cited domains. Wikipedia, NIH, EPA, ScienceDirect are largely clean. The problem lives in the thousands of domains cited once or twice each, and a handful of those had every single cited page flagged as machine-written. One of them describes itself on its own homepage as an AI-powered research tool and notes that accuracy is hard and they're working on it. Four engines cited it 28 times. The mechanism behind this should bother anyone doing this for a living. NewsGuard checked robots.txt across the 500 most engaged news sites. Of the 414 rated high-trust, only 33% allowed all crawlers through. Outlets with a reputation to protect are shutting the door faster than outlets with nothing to lose, and the audit's authors point at research showing exactly that pattern. Bot traffic passed human traffic on web page requests this June. More publishers are going to close. So the retrievable web keeps thinning at the top and thickening at the bottom! One more thing worth knowing. Some attribution setups draft the answer first and go find supporting sources afterward. The research literature calls this post-hoc citation, and one paper describes attributions produced that way as naturally unfaithful. A link under an answer confirms a page exists. It does not confirm the claim came from that page. Keep chasing AI visibility. Just stop treating citation counts as a quality signal. Open the link, find the sentence, and when the sentence isn't on the page, you've learned more than any visibility score will tell you.
-
Leigh McKenzie posted thisA real estate platform went from 10,000 pages to over 425,000 in three months. A transit app went from under 300 pages to more than 10,000. A cloud phone company built a landing page for every area code in the US. A sports betting site added 1,500 glossary entries in six months. A shopping mall added 51,000 pages in a single spring. All of that traffic was earned under rules that have tightened since. Google's scaled content abuse policy has been on the books since March 2024, and enforcement has picked up: spam updates in August 2025, March 2026, and June 2026. The March one finished in under 20 hours, the fastest on record. The sites that held up had one thing in common. Every page carried information that only existed on that page. The transit app's pages cover specific commutes on specific regional lines, with real schedule data behind each one. The area code pages carry actual number availability. The betting glossary is 1,500 definitions people are genuinely searching for. Change the variable on any of those and the answer changes completely. Change the city name in a services template and nothing changes at all. If you want to run this play, this is the sequence I'd use. 1. Start from a dataset, not a keyword export. That decision determines everything downstream. The real estate platform runs on 950 data points per property. The transit app holds live schedules across 300 cities. Ask what data your business already sits on that has never been published as pages. 2. Name the unit in three words or fewer. One page per commute route. One page per area code. One page per betting term. If you can't say it that cleanly, you have a keyword list rather than a page set. 3. Run the swap test before you build the template. Write two sample pages from two real rows and read them side by side. If a proper noun is the only difference between them, stop. Those are doorway pages with better formatting. 4. Ship 150, not 15,000. Later's social media glossary was around 200 pages and reached 60% of site traffic shortly after launch. Wagering Terms spread its 1,500 entries across six months. The 425,000 in three months example is the outlier in this set and carries the highest risk profile in it. Rate of publication is a signal on its own. 5. Build the index page a human would actually use. Every one of these sites has one. An A to Z glossary, a city directory, a route finder. Internal links should flow from there, not from a footer block nobody clicks. 6. Set your pruning rule before launch. Pick a threshold now, something like zero impressions after two full quarters, and hold to it. In a large page set the weakest pages are the liability, and nobody gets around to cleaning them up unless the rule existed from day one. Scaled publishing still works. What changed is how much each individual page has to carry, and that standard applies retroactively to everything you've already indexed.
-
Leigh McKenzie shared thisMost SEO strategies I get sent are activity lists with a strategy label on them. Publish four comparison posts a month. Build ten links from industry publications. Add schema to every URL. Refresh the top 20 pages quarterly. All of that is real work, and every line of it assumes a decision nobody actually made. Six decisions carry a strategy, and each one should be answerable in a single sentence. 1. Visibility. What search has to win you, stated in business terms. If you could hit the target and the bottom line wouldn't feel it, you've written down a metric. 2. Audience. The situation your buyer is in when they start looking, and the order they research in. A buyer who checks Reddit, then G2, then validates on Google and in ChatGPT gives you a completely different brief than one typing a query cold. 3. Territory. The topics you'll compete to own, and the ones you're skipping because a category leader already has them locked. 4. Positioning. A claim that is true, distinct, and visible. Visible is the one teams fail. It means a crawler or an AI system can find proof of it on your site. If the proof only comes out on a sales call, it doesn't exist yet. 5. Presence. What you build on your own site, and what you earn beyond it. Own includes the technical floor and the structure that makes a page retrievable. Earn includes the threads and review sites where your buyers actually ask for recommendations. 6. Measurement. One outcome close to revenue, one leading indicator that moves before it. A dashboard with twelve numbers on it is reporting. Go through those six and see which ones you can answer with a sentence you've already written down somewhere. The ones you'd be making up on the spot are where the strategy is missing, and no amount of publishing covers that gap. Our team put the whole thing into a free workbook, with a coaching prompt you can paste into ChatGPT or Claude if you'd rather talk it through than stare at a blank template: https://lnkd.in/gKia6e2D
-
Leigh McKenzie shared this"Let's build an agent" is the most expensive sentence in SEO right now. The AI requests landing on SEO teams tend to be one of four jobs. The graphic below maps each one to the block it needs. Working through them in order is the whole trick. Sorting a keyword export by intent or drafting meta descriptions is a model and a good prompt. Nothing else. Briefs that depend on your style guide and this week's product docs need retrieval. Live rank data doesn't; call the source directly. Retrieval over a stale export gives you a confident old number. Pushing a fix to the CMS or editing robots.txt means deciding what the model is allowed to do before deciding how it connects. A direct API covers most of it. The agent conversation starts at job four: a workflow that spots a traffic drop, pulls the deploy log, checks whether a noindex shipped, opens a ticket, and waits for a human before anything changes on the site. That one needs orchestration, and it needs the tightest permissions of the four. The production question is where you let the model use judgment, and where the system stays in control through rules, permissions, and approvals. Scope in that order and most SEO projects stop at job two. Visual analogy inspired by: Alex Wang 😃
-
Leigh McKenzie shared thisI’ve been thinking a lot about what it actually means for a brand to “own” a topic in AI search. AI visibility gets much more useful when you stop looking at it as one big score and start asking a narrower question: which topics does your brand actually own? On September 29, I’ll be joining Kevin Indig, Amanda King, and Fernando Angulo for a free Semrush roundtable based on an analysis of six months of ChatGPT data across 1,094 topics. We’ll dig into what separates topic leaders from everyone else, why a strong domain doesn’t automatically translate into visibility for the topics that matter, and where brands have opportunities to build authority next. There’ll be an open Q&A too. 🎟️ Save your spot: https://shr.bi/YeiCgUYP
-
Leigh McKenzie shared thisI sat down with a group of CMOs and VPs at a hedge fund portfolio company this week to talk about AI search, and what stood out was how similar the problems were across very different businesses. Everyone is trying to answer some version of the same questions. Who owns AI visibility? What should we actually measure? How should SEO, content, brand, PR, social, partnerships, and PMM work together? What are we even optimizing toward? I think this is why shared frameworks matter so much right now. Most companies do not need another list of GEO tactics. They need a common way to think about how brand visibility gets built, so different teams can work toward the same outcome. The framework I keep coming back to is: 1. Clarity Does AI understand your brand correctly? 2. Discoverability Can AI systems find and retrieve your content? 3. Authority Does your brand appear qualified to be included? 4. Trust Will AI confidently recommend or act on your brand? The important part is that these layers are connected, but they are not owned by the same team. SEO cannot solve clarity on its own. PR cannot solve discoverability on its own. Brand cannot solve authority on its own. That is why I think AI search needs to be treated more like an operating system for marketing than a new channel somebody has to own. Shared language. Clear ownership. Shared measurement. New North Stars. That is what gives teams a way to build toward the same thing instead of running a bunch of disconnected AI search tactics.
Leigh McKenzie commented on a post
4m
-
Leigh McKenzie reacted on thisLeigh McKenzie reacted on thisSince "ai-instructions" pages seem to be the current AI search hack of the week, if you're going to do it (we'll set aside whether you should for the moment), then at least do it right. Peec AI has a particularly broken implementation. Apparently it's one of their most cited pages. Well yeah, it’s a page stuffed full of keywords and brand facts meant to spoon-feed LLMs. But when a user lands on it... they see... a broken page full of hashtags, asterisks, pipes, and unclickable naked URLs (image 1). Because it's not a markdown page, it's an HTML page masquerading as a markdown page. Peec have nested each "markdown" section inside a code block, inside a paragraph (image 2). Document wise, we have a bunch of code blocks (or paragraphs), with no semantic structure. These are not headings, they're hashtags. They're not lists, they're asterisks. What about when an agent requests markdown? Does that work? No. Because each section is in a <code> block, every block of markdown gets wrapped with backticks (image 3). So even when served as markdown, it's still not proper, valid markdown. Some parsers can probably deal with it. But why not just do it properly? The page, if you must have it, should be an HTML page with regular HTML tags for structure. A human visitor (remember them?) will see a regular page with links they can click and text that doesn't look broken in their browser. Convert it to markdown when requested and that clean HTML will convert to perfectly formatted, valid markdown. The beautiful irony with the current implementation is that it looks like markdown to a human, but doesn't to a machine. End of rant.
-
Leigh McKenzie liked thisLeigh McKenzie liked thisI've just published a new report on The State of AI ✨ in Content Marketing. Some highlights: - 69% of marketers say they're publishing more AI content now than a year ago 📈 - Most respondents say human-written content beats AI, but 25.4% said AI is better (and that number is growing) 👨💻 - The average blog post costs $286, compared with ~$61 for a post created with AI 💰 - One of the fastest-growing, specialist use cases of AI is for keyword clustering 🔑 - 61% of marketers say their view of AI-generated content has become more positive 🤝 A huge thank you to our 301 survey respondents who helped us put this together, plus 8 more highlights in the full report. They let me back on the Ahrefs blog to share this (it’s the first post there) or you can find the link in the comments. Full credit to Ryan Law who pioneered the original version of this report last year, which has been amazing to compare against. Any likes on this are much appreciated, but of course, no pressure from me. Cheers!
-
Leigh McKenzie liked thisLeigh McKenzie liked this3 of my creators got paid $21,750 to fly to SF and rep a brand. That's one of the many ways you can get paid by brands as a creator. Here are a few other ways to not leave money on the table: - charge for exclusivity (15% to 30% of the contract value) - sell usage licenses (30% of post value) - always ask for an affiliate link and push it regularly (I made $20K from one brand like that) - sell your own services through content I shared 5 other ways you can make money on LinkedIn yesterday. Go follow Thibaut Souyris and check my profile to find the post.
-
Leigh McKenzie liked thisLeigh McKenzie liked this'Can we get budget for AI?' is not just one question... Before you sign, place the request on the stack. Each level has one question that decides it. - Individual use: ask which tools are already in use and where customer data is going. This one is a policy call, and it costs nothing to make this week. - Team tools: ask who owns the output and who reviews it before a customer sees it. Then ask what the number looked like before the tool arrived. Without a baseline you will never know if it worked. - Connected workflows: ask whether sales and marketing agree on what a lead is. If they don't, the scoring model will pick a side for them and you will pay for the argument twice. - Customer-facing AI: ask who signed off on what it says, and who checks what it says about you every month. A wrong answer at this level carries your logo. - Agents: ask which single workflow it runs, who can switch it off, and what the pilots error rate was. Anyone who cant answer the third question is asking you to fund a guess. One rule holds across all five: no level gets funded until the one below it has a named owner. Skip a level and the budget compounds the mess along with the return 🫣
Experience & Education
Courses
-
Business Communications
-
-
Data Analytics
-
-
Data Centric Application Development
-
-
Digital Design and Innovation
-
-
Digital Marketing
-
-
Entrepreneurial Marketing
-
-
Financial Accounting
-
Recommendations received
16 people have recommended Leigh
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content