PPC & marketing news, tracked daily.
Paid search, paid social, and feed-driven advertising changes across Google, Microsoft, and the AI ad surfaces.
What changed, as it happens.
Tracked daily from primary sources. The items below are what we're watching for clients right now.
Google Ads' Ad Preview and Diagnosis tool has been stuck behind a captcha for about a week
Google's Ad Preview and Diagnosis tool has been returning a reCAPTCHA "unusual traffic" screen instead of results for roughly a week, which makes it unusable. Search Engine Roundtable reports that Google has given no timeline for a fix.Source: Search Engine Roundtable →
Our take: The tool exists so you can check whether an ad is serving without searching for it yourself and inflating impressions on the keyword you are worried about. With it down, that is exactly the substitute people reach for. The campaign already answers the question without costing you an impression, since a keyword that is serving has impressions against it and a keyword that is not tells you why in its status column.
Read our full take →Four exemptions to the EU's AI labeling rule, written in undefined words
The European Commission set out four exemptions to the Article 50 AI-transparency obligations that became applicable on 2 August. The first covers deepfakes in image, audio, or video that form part of an evidently artistic, creative, satirical, or fictional work; disclosure is still required, but in a manner that does not hamper the display or enjoyment of the work, and the exemption falls away where use is authorized by law to detect, prevent, or investigate criminal offenses. The second covers AI-generated or manipulated text published to inform the public on matters of public interest, provided the content has undergone human review or editorial control and a natural or legal person holds editorial responsibility for the publication. The third covers an AI system performing an assistive function for standard editing, on the condition that it does not substantially alter the input data or its semantics. The fourth removes the disclosure duty for systems that interact directly with people where the interaction is obvious to a natural person who is reasonably well-informed, observant and circumspect. Search Engine Journal notes the Commission has not defined what counts as substantial alteration, what makes an interaction obvious, what an appropriate manner of disclosure looks like, what qualifies as public interest, or what level of human review satisfies the editorial-control condition.Source: Search Engine Journal →
Our take: For a store the tempting exemption is the third one, since most AI use in a catalog is editing help rather than wholesale generation. It will not hold anyone's weight yet. Substantially alter is undefined, and the distance between tightening a description you wrote and handing a model a product name and taking what comes back is exactly the distance the Commission left blank. Exemption two is not the escape hatch it looks like either, because publishing to inform the public on matters of public interest describes journalism, and product copy does not become that because a human reviewed it.
Read our full take →The EU's AI labeling rules bind today, and the ad platforms have not shipped the buttons
Article 50 of the EU AI Act became legally applicable on 2 August 2026, and it splits into two sets of duties. Providers, the companies that build generative systems, must apply machine-readable marking through digitally signed metadata plus an imperceptible watermark for audio, images, and video, watermark free-form text longer than 200 tokens, and offer a free detection tool to deployers, end users, authorities, researchers, and media. Deployers, a category the guidance describes as covering agencies, brands, and publishers, must visibly label deepfakes and AI-generated published text using the EU's icons or equivalent wording reading "AI GENERATED" or "AI MODIFIED". Those labels have to sit at first exposure and stay visible before any overlay covers them, appearing at the start of a video and again after ad breaks, or above or near a headline or in a colophon for text. Penalties reach EUR 15 million or 3% of total worldwide annual turnover for the preceding financial year, whichever is higher. Two further deadlines follow: generative systems already on the market must reach marking conformity by 2 December 2026, and a detection interoperability solution is due by 2 February 2027. Meta signed the EU's Code of Practice on Transparency of AI-Generated Content on 28 July, five days before the obligations bound, having declined the separate General-Purpose AI Code in July 2025 over what it called legal uncertainties. Google signed on 24 July while warning that added regulatory complexity in an unsettled technical field could work against European competitiveness.Source: PPC Land →
Our take: The half that binds an advertiser is the deployer half: visible labels on deepfakes and on AI-generated text you publish, in the EU's icons or equivalent wording. The gap is tooling. Meta's announcement describes principles and forums, and no labeling setting has appeared in Ads Manager to point at, which leaves the obligation sitting with whoever publishes the ad rather than the platform that generated it.
Read our full take →Google Analytics will now tell you when paid clicks are being filed as organic
On 30 July, Google Analytics added a data-quality diagnostic named "Campaign data accuracy is affected by missing URL parameters." It surfaces inside individual reports rather than as a global banner, and it triggers when landing-page URLs arrive carrying a GCLID but missing the GBRAID or gad_ parameters that Google Ads appends. GBRAID, introduced in March 2022, carries campaign data without linking to an individual user and acts as a fallback where GCLID cannot work, such as when a user denies ad_user_data consent; gad_source identifies the ad source and gad_campaignid identifies the campaign. Three mechanisms strip these before the analytics tag fires: redirect chains that drop parameters before the final landing page, CDN or security filters applying canonicalization rules mid-journey, and tags firing before the parameters are readable on the page. Google's documentation also requires gad_ parameters to sit before any hash fragment in the final URL. The loss produces two distinct reporting outcomes. One shows up as "(not set)", which looks broken enough to get investigated. The other credits the paid click to organic search, inflating measured organic performance and suppressing return on ad spend while generating no error anywhere. Every property role can view the diagnostic and Editor or above can act on it, a "View URLs" control lists the affected page paths, and Google says implementation fixes take 24 to 48 hours to register.Source: PPC Land →
Our take: The dangerous outcome is the second one. A paid click filed as organic does not look like an error, it looks like SEO working, and budget drifts toward the channel that did not earn it. If the diagnostic fires in your property, start with the redirect chain on your landing pages, which is where the parameters usually die.
Read our full take →ChatGPT is testing an ad that opens a chatbot built from your website
OpenAI appears to be building a ChatGPT campaign type in which clicking the ad opens a conversation with a business-specific AI agent rather than loading a landing page. The agent is assembled from a business profile ChatGPT generates automatically by crawling the advertiser's website, combined with advertiser-supplied custom instructions and data sources including product feeds, Model Context Protocol tools for live data, and lead-generation forms. It runs on the same infrastructure as Custom GPTs, and inside the conversation it can answer product questions, make recommendations, schedule appointments, troubleshoot, and qualify leads. Entrepreneur Juozas Kaziukenas spotted the capabilities in ChatGPT Ads Manager and posted them on LinkedIn; OpenAI has not announced the format. It is described as reaching a limited group of advertisers, and the shopper-side experience has not been widely observed, so how the ads render and how prominent they are remains unknown.Source: Search Engine Land →
Our take: If this ships, the crawl of your site stops being an SEO concern and becomes ad production. Whatever ChatGPT reads off your pages and your feed is what the agent will say about your products to someone who has already clicked. Thin specs and stale availability turn into answer-quality problems inside a paid conversation you are being billed for.
Read our full take →
Custom ecommerce software, AI apps, and SEO — the work other agencies quote around, built in-house.
20+ years of BigCommerce engineering, now AI-augmented. Tell us your store, your stack, and your deadline — we quote fixed scope on the first call.