Spyglasses Blog

Claude is the easiest AI assistant to leave out of visibility tracking, because every answer is a paid API call. We tested whether the API can stand in for claude.ai on 40 B2B buying questions. Opus 5.5 came close. Sonnet and Haiku named a different set of brands.

Jim Wrubel

10/2/2026

Buyers ask AI who the options are long before they type anybody's brand name. Here's how to map a whole market, see which brands AI actually names, and watch the ranking move every day.

Jim Wrubel

9/18/2026

A ranking of who AI names in a market is also a list of who it barely names. Here's how to read that list for the brands worth calling, and open the pitch with their own numbers instead of your credentials.

Jim Wrubel

9/18/2026

Every market has a story AI keeps repeating, and a short list of publishers it takes that story from. Here's how to write the claims down, track them daily, and find the outlets that set the terms.

Jim Wrubel

9/18/2026

We asked ChatGPT about 48 brands twice a day for ten days and recorded which site it chose to search. It picked the right domain in over 98% of calls in every tier, including brands whose domain has nothing to do with their name. The one thing it doesn't forget is a domain you used to have.

Jim Wrubel

9/14/2026

Sometimes an AI assistant skips the open web and reads one specific site before it answers. Which site it picks tells you who it already trusts, and when it picks an address that looks like yours but isn't, your authority is landing somewhere you don't control. Here's how to read that list weekly and correct the wrong entries.

Jim Wrubel

9/9/2026

Blocking AI crawlers without hurting your AI search visibility comes down to which agents you block, not which companies. Here's how to set a crawler policy on purpose, enforce it where it holds, and confirm you didn't cut off the assistants sending you customers.

Jim Wrubel

9/8/2026

When two companies combine, AI assistants keep answering as if nothing happened. Customers ask whether the acquired product still exists and get last year's answer. Here's how comms teams track both names through announcement, close, and migration, and prove when the market finally caught up.

Jim Wrubel

9/7/2026

Legal and comms want to know when an AI assistant says something harmful about your brand, and where it came from. Social listening won't tell you; AI answers aren't posts and nobody reports them. Here's the monitoring setup that catches a risky claim early, shows how often it surfaces, and traces it back to the sources feeding it.

Jim Wrubel

9/6/2026

When a reporter, an investor, or a candidate asks an AI assistant who your CEO is, the answer comes from whatever the model can find about them. Here's the workflow comms teams use to track an executive as their own subject, measure how accurately AI describes them, compare them against real peers instead of rival companies, and watch the number move.

Jim Wrubel

9/5/2026

Your team shipped 40 pages last quarter and AI visibility went up 6 points. Did the content do that? Most teams can't answer, because nothing connects what they published to what the number did. Here's the setup that ties a specific batch of content to a specific measurable change, using page groups, project goals, and dated release markers.

Jim Wrubel

9/4/2026

How do you measure AI search visibility in a new city, region, or country? Not with your blended brand number, because that's mostly made of the market you already own. Here's the setup agencies use to track a new market on its own line, from the location scope through local competitors, local prompts, local goals, and per-market share of voice.

Jim Wrubel

9/3/2026

A campaign report full of placements says the outreach worked. It doesn't say whether AI assistants picked up the story, repeated your messages, or ever read the outlets you pitched. Here's how PR teams track a campaign in AI answers from kickoff through the retrospective, using the same flight dates and target list they already have.

Jim Wrubel

9/2/2026

New Spyglasses research classifies the sites AI consults directly instead of searching the open web. The list is institutional, high-authority, and there's a specific kind of page that gets you on it.

Jim Wrubel

9/1/2026

Most content updates take two to six weeks to show up in AI answers, because nothing changes until a crawler comes back and re-reads the page. Here's how in-house marketing teams measure their real recrawl cadence, find the pages that never get refetched, and stop scheduling work and reports against a timeline that doesn't exist.

Jim Wrubel

9/1/2026

Two of your pages answer the same question, so AI search can't tell which one to cite and neither one wins. Here's how content teams find overlapping pages, pick the canonical one, and consolidate or differentiate the rest without losing the traffic they already have.

Jim Wrubel

8/31/2026

How do you outrank a competitor in AI answers? Agencies find the exact grounding searches a rival owns in ChatGPT, Gemini, and Perplexity, the pages holding those spots, and the third-party sources feeding them. Here's the quarter-long workflow, step by step.

Jim Wrubel

8/30/2026

You can change the name on your website in an afternoon. AI assistants keep answering with the old one for months, because they read a web that hasn't caught up yet. Here's how comms teams measure the switch and find the sources holding it back.

Jim Wrubel

8/29/2026

Your analytics tool shows you almost none of the traffic hitting your site. Bots don't run JavaScript, so they never appear. Here's how to count them at the server, tell an AI assistant apart from a model trainer, and decide who stays welcome.

Jim Wrubel

8/28/2026

Keyword tools tell you what people type into a search box. AI assistants ask their own questions before they answer, and those questions are a much better content brief. Here's how to collect them, find the ones you never answered, and close the gap.

Jim Wrubel

8/27/2026

Most monthly AI visibility reports are a pile of screenshots assembled the day before the call. Here's how to build one that takes an hour, explains every movement on the chart, and survives the question every client eventually asks.

Jim Wrubel

8/26/2026

A press release gets one shot before it hits the wire. Score the draft against a real AI query while it's still a document on your desk, not after it's synced, syndicated, and indexed with whatever weaknesses it shipped with.

Jim Wrubel

8/11/2026

A new product has no AI visibility history to compare against, so most teams wait until after launch to start tracking it. Set up the property, the prompts, and the competitors before launch day, and the first weeks of AI answers become measurable instead of anecdotal.

Jim Wrubel

8/10/2026

A spelling check catches typos before a draft goes out. This is the same idea for AI citation. Paste a draft in before it's live, score it against the query it should win, and fix what's broken while it's still cheap to fix.

Jim Wrubel

8/9/2026

A cold pitch deck gets skimmed. A report showing exactly where a prospect is losing ground in AI search gets read twice. Here's the four-step workflow for turning a list of domains into a batch of pitch-ready AI visibility reports.

Jim Wrubel

8/8/2026

A media list ranked by reach can still be full of outlets AI never reads. Here's how to score outlets before you pitch, price the placement before you negotiate, and confirm the value after the story runs.

Jim Wrubel

8/7/2026

A full-site audit hands you a list of a hundred things wrong with your pages. Not all of them are stopping AI from citing you. Here's how to tell the difference, fix what actually matters, and prove the fix worked.

Jim Wrubel

8/6/2026

Four days ago we published a bet. Match a synthetic prompt panel to how real buyers actually ask, and it should mirror a human panel completely. We ran the test with the predictions registered in advance. Half the bet paid, and the half that failed turns out to be the more useful result.

Jim Wrubel

8/6/2026

A content brief that says "we need to win this query" starts in a different place than a page you're already trying to fix. Here's the query-first process for picking the right page, scoring it against the exact search AI runs, and revising until it actually ranks.

Jim Wrubel

8/5/2026

A new client signs, or a prospect takes your call, and the first real question is the same either way. How visible is this brand in AI search right now? Here's the five-step playbook for turning a domain name into a baseline that holds up.

Jim Wrubel

8/4/2026

A placement report says the story ran. It doesn't say whether AI ever read it. Here's how to connect a campaign's earned media to what AI assistants actually cite, and report the number that proves the work.

Jim Wrubel

8/2/2026

Critics say AI share of voice from synthetic prompts measures nothing real. Instead of arguing, we ran the test on our own prompt generator, against 143 prompts written by real people, and published every number. What we found changes how you should read your dashboard.

Jim Wrubel

8/2/2026

Your dashboard has a line for paid, organic, email, and social. It probably doesn't have one for AI, so the channel is invisible even when it's working. Here's how to build that line, using the traffic you already have.

Jim Wrubel

8/1/2026

Most of your citation opportunity isn't in content you haven't written yet. It's sitting on your site already, one fixable gate away from getting picked up. A step-by-step playbook for finding those pages, diagnosing why AI skips them, and fixing what's actually broken.

Jim Wrubel

7/31/2026

When a negative story breaks, people don't just read the news. They ask AI about it. A step-by-step playbook for comms teams to find out what AI is saying, trace where it comes from, and measure whether the response is landing.

Jim Wrubel

7/30/2026

A one-time report tells a client where they stand today. Tracking is what proves your work moved the needle. A step-by-step playbook for turning a baseline AI visibility snapshot into a nightly system that catches real movement, not noise.

Jim Wrubel

7/30/2026

The critics of AI visibility tools are right about more than our industry wants to admit. Here's what a defensible version of prompt tracking looks like, question by question.

Jim Wrubel

7/29/2026

We re-ran SparkToro's prompt study with 143 real people's prompts against ChatGPT, daily for a week. The market leaders showed up no matter how people asked. Everyone else's fate depended on the wording. Here's what to do about it.

Jim Wrubel

7/23/2026

If all you're doing is tracking prompts, you're paying to see data you can't act on. Here's what to focus on instead.

Jim Wrubel

7/23/2026

A new Spyglasses study found that Google AI Overviews deep-links to the exact second in a YouTube video 44% of the time. ChatGPT, Gemini, and Perplexity never do. Here's what that means for video in your AI visibility plan.

Jim Wrubel

7/13/2026

Before an AI model ever answers a question about your brand, it's already decided which websites it trusts. That call comes from Common Crawl, and one number predicts it well. Here's what PR and marketing leaders need to know.

Jim Wrubel

7/1/2026

Use AI visibility data to optimize content for AI citations: the 8 gates ChatGPT runs before quoting a page, plus how to test a draft pre-publish.

Jim Wrubel

6/29/2026

We checked the robots.txt policies of the top 51 U.S. news outlets. 61% block AI assistants from citing their content. Here's what that means for your earned media strategy.

Jim Wrubel

3/4/2026

Content refreshes often improve AI readiness and SEO at the same time, but the AI impact usually goes unmeasured. Here's how to change that.

Jim Wrubel

3/2/2026

Good SEO work often improves AI citation readiness as a side effect. Running an AI Readiness Audit before and after proves it.

Jim Wrubel

3/2/2026

Website redesigns can help or hurt AI citation readiness. A scored before/after audit shows which, and gives agencies a deliverable most rivals don't offer.

Jim Wrubel

3/1/2026

Generate prospect-specific AI visibility insights in minutes from a domain name. Turn a cold pitch into a data-backed conversation.

Jim Wrubel

2/28/2026

The gap between AI visibility scores and client results is where agencies lose deals. Two tools close that gap.

Jim Wrubel

2/27/2026

Demonstrating the ROI of PR means more than just tracking share of voice in prompts.

Jim Wrubel

1/12/2026

What if you built an SEO article for a long-tail keyword with 33 Rank Difficulty, and nobody came?

Jim Wrubel

1/12/2026

ChatGPT gets 3.2B visits monthly with 89% never clicking through. Your $100K website redesign optimizes for customers who aren't coming. Here's where they actually are.

Jim Wrubel

1/12/2026

Google's AI Mode has rolled out to all US users and now counts in Search Console data. But the most important metric—how often AI mentions your business without sending clicks—remains invisible.

Jim Wrubel

1/10/2026

11 essential questions to ask yourself to measure if your AI SEO strategy is actually effective.

Jim Wrubel

1/9/2026

What to look for in an AI Visibility Tracking platform for your agency in 2025.

Jim Wrubel

1/8/2026

Traditional conversion metrics miss a huge part of your sales funnel. Learn how AI conversion rates help you optimize for the future of customer discovery.

Jim Wrubel

1/7/2026

With AI models increasingly crawling the web for training data, website owners face a critical decision. Learn when to block AI crawlers and when to welcome them.

Jim Wrubel

1/6/2026

Does adding an LLMs.txt to your site actually do anything? See what our data shows.

Jim Wrubel

1/5/2026

How to Monitor Brand Mentions in ChatGPT, Google AI Overviews, Claude & Perplexity.

Jim Wrubel

1/4/2026