Citation-mania
In the 1630s, the price of a single tulip bulb in Holland could exceed a skilled tradesman’s annual salary, traded on nothing but the promise that whoever bought it next would pay more. Tulip mania measured everyone else’s confidence that the price would keep climbing until, over about a week in February 1637, it didn’t, and the market for tulip bulbs collapsed by ninety percent or more.
I think about this case study in market speculation whenever people get hot to join a bandwagon, as with recent advice on “showing up in AI”. There’s a fast-growing crowd of llms.txt files, AI-visibility subscriptions, and warnings about competitors already “winning” a channel no one can measure reliably. Adopted because … adoption is visibly rising, not because it has been shown to work sustainably over time. This is also concerning because clients asking AI to “optimize” their content will often get dubious recommendations and sources.
How do I show up when someone asks an AI tool for a recommendation?
As the team at SEOPress put it, “SEO isn’t packing its bags, it’s just changing the address.” The answer to “how do I get cited by AI” relies on the same fundamentals that apply to “how do I rank for a search”: who is this content for, and does it genuinely serve them better than what already exists. Most content development work skips that question entirely.
An AI overview can answer a question well enough that nobody ever clicks through to the article it pulled from, so writing something adequate and generic won’t help you. Results from ChatGPT queries carry a particular kind of persuasive weight, because it plays on a natural tendency to confuse cause and correlation. A confident, well-worded AI answer feels like harder evidence than a vague-sounding blog post, but that feeling is mostly manufactured by the specificity of the artifact. This is where creating original, thoughtful, specific content – even if you publish less frequently – can make a bigger difference.
Quality content is harder to automate, because it still requires someone to think critically about a real question rather than just produce at scale.
Since I am always recommending to my clients that they should write blog articles, I wanted to walk through a year of my own content, sorted by what I was interested in exploring.
Disclaimer: This is not rigorous research, just my own experimentation with search optimization techniques as my time and interest allows.
Content, meet Audience
Writing about something because you’re genuinely excited about it is a completely legitimate reason to write. It’s just a different activity, with a different bar for success, than writing to rank, convert, or build authority. You won’t rank for something nobody is searching for, no matter how well you write it, and you won’t convert a reader who was never the intended audience.
In practice, most small business content falls into three groups:
- Content you write because you find it interesting. If inspiration strikes, go for it. But this is mostly for you, it’s not written to rank or convert anyone.
- Content written for people in your field, not your clients. The audience is peers, not prospects. The payoff is reputation inside a professional community, not a phone call.
- Content written to reach and convince a specific buyer. Even here, it matters whether you’re trying to win over a stranger who’s never heard of you, or reinforcing trust with someone already in a conversation with you.
Peer content: solving my own problem in public
A chunk of what I publish isn’t written for clients at all. Posts on WordPress development, ACF block patterns, digging into variable font use – these posts originate from real problems I’m trying to solve where I couldn’t find a definitive resource. The audience is other WordPress developers and being useful to the community I work in is my purpose.
Judged on those merits, these articles are working exactly as intended. Three clicks on 1,052 impressions for “acf blocks” is a fine outcome for a peer-audience post.
Content built to be shared directly, that may also have search value
In November and January, I published two posts – one on phishing and fake domain renewal emails, one on why WordPress updates matter for security. The intent wasn’t to win cold, comparison-shopping searches from strangers. Clients has asked me these exact questions directly, over email and in project calls. I wanted a canonical piece I could point to instead of re-explaining the same warning signs from scratch every time, and something that also reinforces the trust that conversation is built on.
Both posts ended up ranking well anyway. The November post has an average position of 9.5 with 1,950 impressions. The January post is at 8.3 with 570 impressions. Total clicks across both: zero.
Look at what they’re ranking for: “siteground copyright infringement email,” “fake domain renewal,” “‘final notice’ antivirus renewal scam.” These are searches from someone who just got a scary email and needs an answer in the next thirty seconds, not the audience I was writing for.
Out of curiosity, I checked what’s ranking above my results for a couple of those queries: SiteGround’s own official knowledge base, ICANN’s own fraud alert, established security publishers like Malwarebytes already covering the same scam campaign. I was never going to win that search.
Which points at something the ranking data can’t tell me either way: whether the posts are fulfilling their purpose. Have they made a real conversation easier? Saved me from re-explaining the same thing over email? Given someone something reassuring to forward along? That’s not a Search Console metric, and it’s not something impressions or position can confirm or deny. The job of these articles is to support existing relationships of trust, not to manufacture it out of a cold search.
Buyer-aware content that did the research: my “site maintenance” page
My WordPress maintenance page has existed since 2021, doing fine, unremarkable work. Last fall I noticed it was pulling far more impressions than my other service pages but almost none of that visibility was becoming inquiries. So this time, before rewriting anything, I compared it to pages that were ranking well for the same terms and asked what they had that mine didn’t.
The gap was specificity and comprehensive breadth (not necessarily good or useful writing, alas). My content was vague where theirs was concrete. So in October and November of 2025, I rewrote it: adding client testimonials, refining the offer, rebuilding the FAQ around client questions and the “people also ask” prompts Google surfaces for that exact search.
Nine months later, the rankings side of this looks like the earlier examples: average position climbed from the 40s and 50s to the 13–14 range, weekly impressions grew from under 500 to a peak near 4,800, and total clicks across the whole twelve months landed at five. Judged purely on Search Console, this would read as another wash.
When a prompt requires current, specific, or verifiable information – which includes most commercial and research queries – LLMs like ChatGPT, Perplexity, and Gemini retrieve relevant documents from external search indexes before generating their response. For these queries, the retrieval step is key: if your content isn’t indexed and ranking in the search engine, it cannot enter the model’s context window.
Lily Ray, Your GEO Strategy Might Be Destroying Your SEO
In the spring and summer of 2026, three new maintenance inquiries came in from people who, when asked how they found me, said some version of “I asked ChatGPT.” They couldn’t point to an exact page they were referred to, just that they’d asked an AI tool for a recommendation and gotten mine.
It’s tempting to draw a straight line from “I rewrote the page in October” to “ChatGPT recommended me in June”. But is that true or truthy?
Pattern v. Coincidence

A general AI-visibility tracker I run through SE Ranking climbed from roughly 10–15% in late April to a sustained 50–75% through most of June, which is close to the window those three inquiries arrived in. That’s a percentage across a small set of tracked prompts, measured daily, which is exactly the kind of series where a good week can look like a trend.
Rand Fishkin’s research on AI-driven search makes a relevant point – these tools give inconsistent answers to the same question asked twice, and there’s rarely a reliable way to trace one recommendation back to one piece of content.
To get mathematical about it, there’s a <1 in 100 chance that ChatGPT or Google’s AI, if asked 100X, will give you the same list of brands in any two responses. Claude is just slightly more likely to give you the same list twice in a hundred runs, but even less likely to do so in the same order.
Rand Fishkin, SparkToro.com – Ais are highly inconsistent
Fishkin’s argument predicts exactly this kind of result: he’s said full-funnel, page-level attribution doesn’t really work anymore for AI-driven visibility, and the best move is tracking aggregate brand presence instead of chasing which single URL gets the credit.
The other pattern is publishing volume. From September 2025 through April 2026, I put out 18 posts spanning WordPress, web development, content strategy, SEO, UX, accessibility. Only a handful touched security or AI specifically. I shared my articles on LinkedIn and BlueSky. Lot’s of LinkedIn traffic referrals, mostly from the peer-driven topics.
If there’s a finding in this, showing up consistently, across a normal range of real topics, seemed to matter more than picking one narrow niche and hammering it. That’s a more realistic goal for most small business owners than an aggressive content marketing strategy that demands multiples posts per week paired with social media campaigns.
In writing articles, I also wanted to maintain a standard of quality that mattered to me. I wanted my work to be moderately creative and personally satisfying. If you can match the specificity and depth of what’s ranking, and also make something that is “good” (human, genuine, original, well-written), it should perform better over time.
Before you publish, make sure you ask the right questions
If your process right now is “someone writes something, I glance at it, we hit publish,” here’s an an adjustment which will increase your potential benefits that costs maybe thirty extra minutes:
- What is the purpose of your content (before writing or prompting)? Enthusiasm content you create for yourself, peer content, and buyer content have different jobs and different success metrics.
- Who’s going to read this? A stranger, or someone already talking to you? If it’s a stranger finding you cold, find the exact question they’d search, then check who already owns it – other businesses like yours, or a name-brand a worried person would already trust more, like a bank, a host, or a government site? If it’s the second one, that’s important to know before you spend an afternoon competing for a click you were never going to get (in this case, explore a longer tail keyphrase).
- Check your own site before creating something new. Improving an existing page can be a better use of limited time than adding another thin post to a pile.
- Answer the question in the first two sentences. Save context for after. This is what makes a section quotable, to a human skimming and to an AI system deciding what to cite.
- When you’re the one editing, check for specificity, not just tone. Ask “would this paragraph describe five other companies or is it unique to us – does it have a point of view or strong opinion?”
- Don’t wait for a perfect content plan. My own results came from eight months of ordinary, unglamorous publishing, not an engineered content cluster. Pick a bucket, pick a real question, and keep the habit going.
Work in Progress
A lot of real content is blended. An FAQ can serve both a worried client and a curious peer at once, and the security posts are arguably evidence of that: written with buyer intent, but the query behavior that got exposed (panic-searching, trust-brand competition) suggests the actual audience search intent was closer to “verify this isn’t a scam” than “evaluate a WordPress maintenance provider.” That could be treated as a missed research step, which is true, but it’s also a reminder that intent isn’t fully something a writer decides.
Searcher intent is the thing that governs the outcome of your search optimization work, not tricks to exploit short‑term loopholes. Research and experimentation show you what intent your audience has, how to satisfy it with the right content and format, and what moves your metrics over time.
Image: A Satire of Tulip Mania by Jan Brueghel the Younger (c. 1640) depicts speculators as brainless monkeys in contemporary upper-class dress. In a commentary on the economic folly, one monkey urinates on the previously valuable plants, others appear in debtor’s court and one is carried to the grave. Frans Hals Museum, Haarlem, Netherlands., Public Domain














