The GEO Show, Ep. 5: GPTBot Read Zero of Your JavaScript Pages

GEOforge · The GEO Show, Episode 5

GPTBot read zero of your JavaScript pages in a 41-day crawl test. Google found every one of them.

By Paris ChildressGEOforge9 min read

Eleven stories dropped in the GEO world this week, and the sharpest one should worry anyone running a JavaScript-heavy site: AI crawlers read zero of the pages Google indexed without any trouble at all.

We cover all of it on Episode 5 of The GEO Show. Here’s the roundup, with the receipts.

ChatGPT Ads just made attribution messier

OpenAI added view-through conversion reporting to ChatGPT Ads, with a one-day window: if someone sees but doesn’t click an ad and converts within 24 hours, OpenAI now counts it. Anyone who has run paid social recognizes this measurement stack immediately.

The catch is that a single lead can now be claimed three times over: a last-click conversion in Google Ads, a view-through conversion in Meta or LinkedIn, and a view-through conversion in ChatGPT Ads, all from the same person. Until platforms sort out de-duplication, the fix is manual. Add “how did you hear about us” to every lead form, with “AI chatbot” as an option. It won’t be perfect, but it’s directionally honest in a way your ad platforms currently aren’t.

GEO is becoming a job title at Cisco and Replit

Cisco Careers posted an opening for an “SEO/GEO Content Strategist.” Replit is recruiting a lead role spanning GEO and SEO, tied directly to sign-ups and revenue rather than traffic.

Large enterprises are starting to insource GEO the same way they once insourced SEO. If your business runs on agency engagements with companies this size, that’s worth watching now, not after the first client tells you they’re building the function in-house.

AI crawlers found zero of your JavaScript pages

A 41-day field test across roughly 1,000 pages logged 30,180 bot requests. Googlebot and Google-Other crawled JavaScript-injected links without issue. GPTBot, ClaudeBot, Perplexity, and every other tested AI crawler read zero of the pages accessible only through those links. Once the same links were converted to plain HTML, GPTBot discovered 250 pages, the first within 48 hours.

Google spent years building a rendering pipeline that executes JavaScript before indexing a page. AI crawlers were not built the same way, and there’s no sign yet that they’re racing to catch up. That means a page can be fully indexed by Google and completely invisible to every LLM at the same time, and nothing in your Search Console data will warn you.

The fix isn’t abandoning JavaScript. It’s making sure a static HTML version of the same content, ideally with structured data, exists alongside it so LLM bots have something to actually crawl.

A new entrant is merging monitoring and execution

Surva.ai now spans ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews in a single platform: visibility tracking, competitor analysis, answer gaps, organic search data, content generation, and referral measurement together.

Read-only monitoring tools were already becoming commoditized, with some offered for free. Now the fully integrated monitor-diagnose-draft workflow is heading the same direction. Tracking your share of voice was never the hard part. Doing something about it is, and that’s where the category is consolidating.

Should content briefs be built around tasks instead of keywords?

Google’s own AI Mode data shows the average query is three times longer than a traditional search query. More than one in six US searches now use voice or images. Planning-related AI Mode queries have grown 80% faster than AI Mode queries overall, and brainstorming queries have grown 30% faster.

Content briefs need to model the whole journey, exploring, comparing, validating, implementing, not a single landing page built for one keyword. That’s a real shift in how content gets planned, not just how it gets written.

Is the real GEO opportunity smaller than the keyword universe?

Duane Forrester of Search Engine Journal argues that AI systems collapse many phrasings of the same need into a small consideration set, which makes “is this category still contestable” a more useful question than tracking dozens of keyword variants. He discloses a commercial interest through Citation IQ, so we’re treating this as a hypothesis rather than settled evidence.

There’s still no tool that reports actual prompt volume the way keyword tools report search volume. Keyword volume remains the best available proxy, and this research is a reason to hold that proxy loosely rather than convert it one-to-one into prompt counts.

Interactive tools are still worth building

StudioHawk’s Lawrence Hitchens analyzed 1.2 million AI referral sessions across more than 600 businesses. Interactive tools, templates, and calculators earned 7.5 times their share of AI referral traffic relative to how many such pages existed. How-to content, definitions, comparison pages, and guides also over-indexed. Standard blog posts underperformed.

Despite Google building its own tools directly into AI Overviews and AI Mode, standalone interactive tools remain one of the strongest investments in this data, and building one has never been easier.

Gemini’s citations are wildly unstable

Steady Demand analyzed 14,472 citations from 1,487 local queries across 50 US metros. Nearly 60% of Gemini citations pointed to businesses’ own websites, but repeated identical queries shared citation sources only about 40% of the time. Google returned the same top business just 7% of the time in AI Overviews, versus 90% for the classic local pack. Gemini and ChatGPT cited the same domains only 8% of the time.

Ranking well in the local pack is no guarantee of AI citation. Treat AI visibility as a probability distribution, not a fixed position you earn once and keep.

Rising Search Console impressions can mean nothing

Ivris Tech reports two pages that accumulated 542,615 Google impressions at an average position of 3.6, and only seven clicks between March 19th and August 11th. Queries of seven words or more generated 414,350 impressions and eight clicks. The authors are explicit that Search Console can’t prove those impressions came from humans rather than AI systems, synthetic agents, or rank trackers, and it can’t rule that out either.

Rising impressions stopped being a trustworthy standalone KPI a while ago. This is the clearest evidence yet of why.

GEO audits are commoditizing fast

Romanian startup Human AI Labs launched Blindspot, which scores websites against 70 deterministic criteria spanning crawler access, structured data, semantic clarity, and trust signals. Beta audits averaged 48 out of 100, and the company hasn’t independently shown those scores cause higher AI visibility.

The same thing happened to SEO audits years ago. They became hard to sell as a standalone service and turned into the on-ramp for a long-term retainer instead. GEO audits appear to be heading the same way, and they can now be done thoroughly, almost instantly, with AI.

EU AI Act transparency rules are live, not theoretical

Article 50 of the EU AI Act has applied since August 2nd, 2026. Providers of generative systems face machine-readable marking requirements. Deployers must disclose certain deepfakes and AI-generated text on matters of public interest when it lacks human review or editorial control. Penalties can reach €15 million or 3% of worldwide annual turnover.

For a company the size of Anthropic, whose annual recurring revenue run rate is reportedly approaching $70 billion, 3% is not a theoretical number. Content that focuses on information gain and proprietary knowledge from the brand itself, regardless of how much AI assisted in producing it, is unlikely to face suppression under these rules.

The bottom line

Four moves to make out of this week’s eleven stories:

• Audit whether GPTBot, ClaudeBot, and PerplexityBot can actually read your JavaScript-dependent pages, not just whether Google can. Pull server logs and check, don’t assume.
• Build content briefs around the whole buyer journey, not one keyword. AI Mode queries are already three times longer than a traditional search.
• Stop trusting rising Search Console impressions on their own. A metric that can’t distinguish bots from buyers isn’t a health signal anymore.
• Watch the category shift from monitoring to execution. Read-only tracking is commoditizing fast, GEO audits included.

Sources & method. ChatGPT Ads view-through conversion reporting per OpenAI’s product update. Cisco Careers and Replit job postings, publicly listed. JavaScript crawler field test: 41 days, roughly 1,000 pages, 30,180 logged bot requests. Surva.ai platform capabilities per its published documentation. AI Mode query-length and growth data from Google’s own published analysis. GEO opportunity-sizing argument from Duane Forrester, Search Engine Journal, who discloses a commercial interest through Citation IQ. Interactive tool referral data from StudioHawk’s analysis of 1.2 million AI referral sessions across 600+ businesses. Gemini citation stability data from Steady Demand’s analysis of 14,472 citations across 50 US metros. Search Console impression data from Ivris Tech, covering March 19 to August 11, 2026. Blindspot audit tool and beta scoring data from Human AI Labs. EU AI Act Article 50 requirements and penalties per the regulation’s text, in force since August 2, 2026. Full breakdown and discussion in Episode 5 of The GEO Show.