LabraSEO

Ads land in AI Mode with Personal Intelligence 🤖, Google AI Mode lets you pick preferred sources 📋, AI tools cite a made-up concept in a week 🕵️

Issue 18, 20 October 2026

What changed in search and AI answers, for marketing teams, in five minutes.

Headlines & Launches

Google starts showing ads in AI Mode when Personal Intelligence is on (4 minute read)

Google confirmed it is testing ads in AI Mode for users who connect their apps through Personal Intelligence, reversing a promise made in March that this mode would stay ad-free. Google told Search Engine Land that ads will relate to the query, the response context and the user's interests, though not directly to Gmail or Photos content. If you run ads, expect a new placement to test; if you rely on organic visibility in AI Mode, this is one more area where paid results may compete for attention.

Google AI Mode may let users manage preferred sources from citations (2 minute read)

Google is testing a "Source preferences" button inside the AI Mode citations sidebar that opens the preferred sources list so users can add or remove sites, spotted by Damien Andell and reported by Barry Schwartz. AI Mode has supported preferred sources since May, but surfacing the control at the point of citation makes it easier for users to shape which sites they see repeatedly. If this rolls out widely, being a site a reader actively chooses to prefer becomes a visibility lever worth tracking.

OpenAI adds visual ad formats and weekly reports to ChatGPT Ads (3 minute read)

OpenAI rolled out wider and full-width image cards for ChatGPT ads, a redesigned conversions overview, and new weekly performance emails for advertisers, according to an email reported by Barry Schwartz. OpenAI says the new ad format tests use existing campaigns and creative with no advertiser action required. If you advertise in ChatGPT, check your account for the new conversions tab and weekly summaries, since ad formats there are evolving fast.

DeepSeek opens its search engine to developers with a GEO filter (4 minute read)

DeepSeek launched a Search API on 17 October that it says filters SEO spam and "GEO manipulation," and every result includes a page-level authority tier and relevance score alongside site: query support. DeepSeek has not defined what counts as GEO manipulation or explained its detection method, and its privacy policy suggests queries may still flow through a third-party search supplier. If your traffic includes Chinese-language or DeepSeek audiences, this is a new index worth checking, though how it treats optimised content is still unclear.

Deep Dives & Analysis

AI answer engines picked up a made-up concept within a week (5 minute read)

Fudan University researchers posted articles about an invented concept on nine sites that ChatGPT, Perplexity, Grok and Google AI already cite often, and all four tools cited the posts within seven days, while the same tools mostly ignored posts about a made-up product or an opinion topic on the same sites. Six to ten of each tool's ten most-cited sources accept posts from ordinary users, with Medium, GitHub, LinkedIn and Facebook rated easy to join. This suggests publishing on frequently-cited, open platforms can get new claims into AI answers fast, which cuts both ways for brand visibility and for the risk of misinformation.

Meta narrows its crawl clause after publisher pushback (4 minute read)

Meta removed "training" from the list of uses in its upcoming Commercial Terms crawling clause, which takes effect 30 October, following objections from the European Publishers Council and the UK's News Media Association and two Nieman Lab reports. The revised clause ties crawling to advertising and business products rather than general AI training, though legal experts say the wording could still be read broadly, and the permission to crawl business-linked websites remains in place. If you run a Facebook Page without buying ads, Meta's crawling terms still apply to you, so this is worth a read before the change takes effect.

Tactics & Tools

RankZero's free AI SEO audit (Sponsor)

See whether ChatGPT, Perplexity and Google AI Overviews can find and cite your brand. Free, results in 30 seconds, no signup.

From X and LinkedIn

Glenn Gabe: Check whether your JavaScript content renders for AI search bots (1 minute read)

Glenn Gabe retested AI search platforms' ability to render JavaScript-based content over a year after his first test and found ChatGPT, Claude and Perplexity still struggle with it, while Gemini and Copilot handle it better because they lean on Google's and Bing's existing indexes. If key content on your pages depends on JavaScript to appear, most AI search tools may not see it; server-side rendering remains the safer choice for anything you want cited.

Lily Ray: ChatGPT is adding disclaimers about who writes the rankings it cites (1 minute read)

Lily Ray tracked around 700 ChatGPT answers to B2B software prompts and found 82 included a disclaimer in the past week, with 19 specifically warning that vendors, affiliated publishers or paid placements may have shaped the rankings and reviews ChatGPT draws on. If you publish comparison content or reviews, expect ChatGPT to flag self-interest more often, so disclosing affiliations and methodology clearly may matter more for how your content is treated.

Aleyda Solis: rising search volume for a commercial query still meant fewer organic clicks (1 minute read)

Aleyda Solis found that for the query "college backpacks" in the US, search volume rose 20% year over year but organic clicks fell about 6% while paid clicks rose 31%, as the SERP added an AI Overview, a second shopping pack and People Also Also Ask results and dropped two organic slots. It is a reminder that more demand for your category does not guarantee more organic traffic if the page is now more crowded with paid and AI features.