Art of Photography AI

Found, Not Chosen

Draws the line between discovery -- whether AI engines can resolve you, name in and correct facts out -- and distinguishability -- whether they choose you, a buyer's question in and your name out. Michael appears in five of five engines for a mobile photography studio in Tucson but zero for fine art pet portraits, where several recommend painters instead. Argues recommendation begins with classification: when a query names the profession the engines find him, when it names the product they build a different candidate set. Covers an incognito ChatGPT session that read a consolidated website as diluted specialization, the deliberate case for keeping shelter work and executive portraiture under one entity, and a fabricated JSON-LD URL (/store/saguaro-collection) that an AI-assisted schema draft planted in his own markup and Googlebot crawled for months. The through-line: hallucination did not disappear, it moved into interpretation, classification, and infrastructure, so verification belongs inside the work as a scheduled habit rather than a setup task. Paid.

The Address Was Real. It Just Wasn’t Mine.

Opens with Gemini inventing a business address in generated JSON-LD -- a real Tucson address that belongs to a restaurant, not the studio -- and builds the fix: an entity card, one dated document holding legal name, founding date, credentials, and service categories, plus written audience profiles that capture decision drivers rather than demographics. The unexpected payoff was auditing: checked against those documents, two existing site pages surfaced errors a normal proofread would never catch, including a compressed biography error on a page an AI helped write. The argument: AI needs something written down and dated to check against, because the next error will not be as obvious as your own address. Free.

The Report Calls It Wasted Time. I Call It Tuition.

Reads BCG's "AI brain fry" research and Glean's Work AI Index against a year of running five AI tools in a working photography business. The surveys put the hidden cost of AI at roughly 6.5 hours a week of "bot sitting" -- feeding context, watching output, debugging, cleaning up -- and find 40% of workers shipping AI work they could not explain if asked. Michael argues that exhaustion-driven satisficing is the generic photographer seen from the inside, and that the same hours read as tuition rather than waste when the curiosity survives. Paid.

Eight Percent

Traces a suspicious share-of-voice number in HubSpot's free AEO grader to its source: a hardcoded fallback of eight percent, written into the code for whenever the model has no real answer. Documents rebuilding the measurement tool twice -- the first rebuild lied in exactly the way the original had, the second stopped lying at the price of watching most scores fall. The core argument: when a tool has nothing, it should say it has nothing. Paid.

The Grader I Started Using Instead

Points the rebuilt AEO grader at a brand with almost no signal -- this newsletter, three months old -- and gets near-consensus low scores from Perplexity, OpenAI, Gemini, Grok, and Claude, where the mature photography brand had pulled fifty points of disagreement. Argues that free AEO evaluation tools are optimized sales funnels, and shows what an honest grader reports when the signal genuinely is not there yet. Paid.

Memory That Doesn’t Ask

OpenAI's Dreaming memory update synthesizes context from chat history without being asked, and for anyone running client work through one account that is a confidentiality question, not a convenience story. The failure mode is not plausible fiction: the detail that crosses between clients is real, just real about the wrong client, so nothing reads as inaccurate while the context is the breach. Covers the defenses -- separate governed spaces, Temporary Chat, business-tier subscriptions, and reviewing what the tool thinks it knows. Free.

What the Tool Found on a Mature Brand

Runs seven subspecialty queries through the AEO grader against Michael Kloth Photography -- twenty years in business, two published books, PPA and ASMP membership, registry-level signal -- expecting uniformly high scores, and gets a spread from 28 to 78. Argues that the scatter is the actual information: once you understand why a brand with real signal scatters across fifty points, you stop reading the headline number as your AI visibility and start reading what is underneath it. Paid.

The Wall Visualizer Works Now

Volume 3 closes the wall visualizer thread: the broken Send to Michael function turned out to be two compounded bugs -- an unregistered endpoint and a panel re-trigger hiding the contact form -- fixed in one session with the Claude Mythos preview after two months of intermittent frustration. Also covers reworking print sizing to snap to standard print sizes and report physical dimensions instead of pixels. The tool is live at visualizer.michaelklothphotography.com. Free.

It’s Now in the Documentation

Google’s June 5, 2026 developer updates formally recognize “optimizing for generative AI” (AEO/GEO) as a hireable discipline while warning that third-party tools lack ranking data and can’t reliably predict performance; this free issue frames a three-part paid arc where Michael stress-tests HubSpot’s AEO Grader on both a mature photography brand and a three‑month‑old newsletter, exposes funnel-driven flaws in its architecture, and builds a more honest way to measure real AI visibility.

The Queries, The Prompts, and What To Do With What Comes Back

Lays out a practical method for AI‑era market research: run audience research prompts across tools like Gemini, Perplexity, Claude, ChatGPT, Grok, and Copilot to surface real client questions and vocabulary, then use client‑simulation prompts in incognito to see what those same tools actually say about your business, turning the results into a living content gap map and a ten‑minute monthly site and brand audit.

What Were You Doing?

Uses real Gemini queries phrased in client language to show how AI assistants synthesize information about a photography business, arguing that photographers should regularly run similar prompts across multiple tools to see what’s being said about them, capture true client vocabulary, and track how those synthesized answers change over time.

What to Keep Out

Applies Brian Halligan’s “strategic illegibility” idea to a solo photography practice, distinguishing between the AI‑readable layer (voice, process, credentials, prohibitions) and the tacit, embodied layer of expertise that lives in judgment and pattern recognition, and arguing that some knowledge should stay undocumented because writing it down would make it smaller than the real thing.

The Patience Transfer

Shows how twenty years of shelter photography built a patience skill set that transfers directly into executive portrait work, where subjects often perform being photographed instead of actually being present, and connects that same lived experience to AI voice infrastructure, which only works when there is a real, hard‑earned voice for it to preserve.

What Schema Actually Does (And What It Doesn’t)

Honest accounting of what @graph JSON-LD schema markup does and does not do for AI search visibility. Confirmed to help Google and Bing ground their generative AI systems; unconfirmed for ChatGPT, Claude, and Perplexity. Covers the implementation built on michaelklothphotography.com via Squarespace Code Injection and the key limitation: schema makes first-party content more legible, but third-party mentions still dominate AI citation signals. Paid.

Google Just Said the Quiet Part Out Loud

Danny Sullivan's April 2026 Google Search Central Toronto presentation declared commodity content finished. Michael runs Kipp Bodnar's six-checkpoint non-commodity content grader against real pages on michaelklothphotography.com -- scores of 21 (blog index, a grader limitation), 69, and 74 -- and identifies where the framework applies consequentially (blog posts, informational queries) versus where it does not (service pages doing navigational and transactional work). Key finding: the gap is a structural strategy decision, not a writing failure. Free.

From LinkedIn Comment to Speaking Relationship

How a LinkedIn comment from Kristen Hassen (Outcomes Consulting, former PACC executive director) led to a guest slot in her AI for Animal Welfare Professionals course, a peer-exchange Zoom, and a speaking engagement at her First Fridays open call for animal welfare professionals. Covers the presentation built with NotebookLM and Claude, hands-on Gemini photo editing experiments run the night before the call, and the argument framed for shelter organizations: operational pragmatism over uncritical adoption or reflexive opposition. PDF of presentation deck attached free to all subscribers. Paid.

Who Does the AI Think You Are?

Michael runs the same prompt across Claude, ChatGPT, Gemini, and Perplexity: what sources do you treat as authoritative when someone asks about professional photography, and why? Methodology is explicitly replicable. The spread of responses -- including each system's varying ability to accurately describe its own sourcing behavior -- is treated as the research. Paid.

The Client the AI Already Knows

The invisible prompt concept extended into implications for photographers. AI tools with persistent user context are matching clients to photographers against accumulated portraits of who those clients are, not just what they typed in the current session. Covers pay-to-play AI advertising (enterprise-scale only, irrelevant to working photographers), the finding that roughly 85% of AI brand citations come from third-party sources rather than brand-owned content, and the argument that authentic signal built honestly over time compounds in ways that optimization tactics do not. Free.

Building the Guide That Makes Everything Else Work

The actual methodology for building voice documentation: specific components, where the material comes from, how to feed it to the tool, and what to do when the output still does not sound right. Companion to "You Need to Hear About This AI Thing," which framed the why; this issue covers the how. Paid.

Vibe Coding a Wall Visualizer, Part Two

Part 2 of 2. PhotoShelter API limitations, the CORS problem that was not a CORS problem (hours spent diagnosing a confident wrong answer), what the screenshot-and-iterate loop actually feels like, and an outstanding bug -- the contact form that should appear after image download does not. Honest about what sustained work with these tools gives you and what it does not. Free.

I Spent a Day Vibe Coding a Room Visualizer for My Photography Business. Here’s What I Learned Before Testing a Single Line of It.

Part 1 of 2. Building a wall visualizer web application using AI coding agents without an engineering background. Introduces the Dag/Norbert two-agent workflow adapted from Dan Roth's builder/reviewer setup. Covers the Squarespace/PhotoShelter API constraint that pushed the app to a WordPress subdomain on InMotion Hosting, and what agentic work actually looks like for a solo business owner. Free.

You Need to Hear About This AI Thing

Introduction to AI tools for photography business use, framed around the gap between generic AI output and output that sounds like you. Opens with a deliberately terrible AI-generated session confirmation email. Covers the main tools, free vs. paid tier tradeoffs, data privacy considerations, and why building voice documentation is the work, not the setup. Free.

Issue 1: How I Got Here, and Why That Matters Now

Origin story. Michael's arc from cancer biology research to photography, MFA at Academy of Art University, two books published by Merrell Publishers (Shelter Cats, Shelter Puppies), and 20 years of professional practice across three states. Frames the AI moment against two prior disruptions -- digital SLRs and the iPhone -- and argues that photographers who understood each technology built durable businesses while those who waited caught up at a disadvantage. Free.

Welcome to the Art & Business of Photography in the Era of AI

Teaser post. Introduces the newsletter's premise via an AI-altered image of Michael's dogs Chica and Birdy in superhero costumes: an MFA-trained photographer using AI tools seriously in a real photography business. Free.

Coming March 10, 2026 @ 9:00 AM

Pre-launch announcement of the newsletter's March 10, 2026 debut: twenty years of professional photography, an MFA, and a working business meeting the AI era. Free.