TRANSCRIPT 🔗
WEBVTT
00:00:01.280 --> 00:00:02.879
Hey, it's Will Scott's VoiceClone.
00:00:02.959 --> 00:00:04.799
The content is curated and reviewed by Will.
00:00:04.879 --> 00:00:05.679
The voice is AI.
00:00:05.839 --> 00:00:09.439
Welcome to the Search Influence weekly briefing for Tuesday, October 20th, 2026.
00:00:09.599 --> 00:00:15.759
This week, the platforms gave marketers more links, more ad controls, more personalized placements, and more permissions to read business websites.
00:00:15.919 --> 00:00:16.879
That sounds like progress.
00:00:17.039 --> 00:00:19.679
The catch is that the measurement still lags behind the product.
00:00:19.839 --> 00:00:23.920
Google can put a source link higher in an AI answer without telling us whether anyone clicked it.
00:00:24.079 --> 00:00:27.039
A budget setting can look like a cap, even when Google calls it best effort.
00:00:27.199 --> 00:00:30.079
A referral source can disappear from analytics while the people keep arriving.
00:00:30.239 --> 00:00:33.920
The practical theme is to check the outcome before we call the product change an improvement.
00:00:34.079 --> 00:00:35.039
Six items made the cut.
00:00:35.200 --> 00:00:40.560
Google is moving more links higher in AI mode and AI overviews, but it still has not added the click data publishers want.
00:00:40.719 --> 00:00:45.920
ChatGPT's referral tags disappeared from many observed links, which may have pushed real visits into direct traffic.
00:00:46.000 --> 00:00:48.240
The tags later returned and most of Profound's checks.
00:00:48.399 --> 00:00:52.399
Google ads added channel priorities and percentage allocations for lower priority inventory.
00:00:52.560 --> 00:00:54.640
Those percentages are not guaranteed hardcaps.
00:00:54.799 --> 00:00:59.359
Google also started testing ads for AI mode users who connect apps through personal intelligence.
00:00:59.520 --> 00:01:05.359
Meta launched a reservation-style ad product called AdSuperburst for concentrated 3 to 24 hour campaigns.
00:01:05.439 --> 00:01:12.879
And Meta's new commercial terms authorize crawling of websites associated with business accounts, while current English language editions have not all shown the same wording.
00:01:13.040 --> 00:01:20.319
The expert panel review adds context from Barry Schwartz, Anuet Bola, Cedric Husler, Profound, Mark Ritson, and legal experts quoted in the Meta reporting.
00:01:20.480 --> 00:01:22.879
Google and Meta's own pages establish the product facts.
00:01:22.959 --> 00:01:26.640
The experts help us understand what is missing, what is still uncertain, and what we should do next.
00:01:26.799 --> 00:01:29.599
Start with Google's effort to put more of the web back into AI search.
00:01:29.760 --> 00:01:35.680
Google says it is increasing the number of inline links in AI mode and AI overviews and showing them higher in responses more often.
00:01:35.760 --> 00:01:41.680
It is also testing a dedicated section near the top for one or two official links when a person needs the exact place to complete a task.
00:01:41.840 --> 00:01:43.920
There are three more changes in the same announcement.
00:01:44.079 --> 00:01:47.040
Timely article carousels are expanding to more locations and languages.
00:01:47.200 --> 00:01:50.959
Links at the end of some AI answers are getting larger designs and image thumbnails.
00:01:51.120 --> 00:01:55.120
Google is also testing clickable creator and publisher names on video results in AI mode.
00:01:55.280 --> 00:01:57.439
Those names lead to a search profile on Google.
00:01:57.599 --> 00:01:59.840
The opportunity is clearest for an official task page.
00:02:00.000 --> 00:02:04.239
If somebody wants to schedule an appointment, apply, book a room, check pricing, or request information.
00:02:04.400 --> 00:02:07.680
Google says it may highlight one or two official destinations near the top of the answer.
00:02:07.840 --> 00:02:09.520
Barry Schwartz points to the missing piece.
00:02:09.680 --> 00:02:12.080
Google did not publish click results for any of these layouts.
00:02:12.240 --> 00:02:19.199
When he asked about click data in the AI performance reports, Google said it is still working with website owners on which metrics would be helpful and will add more over time.
00:02:19.360 --> 00:02:21.280
A citation and a visit are different outcomes.
00:02:21.439 --> 00:02:26.479
A bigger link might send more traffic or it might make the source look better inside an answer that already satisfied the user.
00:02:26.639 --> 00:02:28.240
We will not know from the announcement alone.
00:02:28.400 --> 00:02:31.280
For clients, review the pages that complete a real customer task.
00:02:31.439 --> 00:02:34.639
Make the purpose, organization, location, and action unmistakable.
00:02:34.719 --> 00:02:42.560
Then test the priority queries and record whether Google cites the page where the link appears, whether landing page traffic changes, and whether those visits turn into leads or sales.
00:02:42.719 --> 00:02:45.039
The second story is the ChatGPT attribution gap.
00:02:45.199 --> 00:02:53.360
During October 8th through the 16th, researchers and marketers noticed that many ChatGPT links were missing the UTM tag that normally identifies ChatGPT as the source.
00:02:53.439 --> 00:02:56.319
In some cases, the browser also passed no usable referrer.
00:02:56.400 --> 00:03:00.240
Analytics had no source to classify, so a real ChatGPT visit could appear as direct.
00:03:00.400 --> 00:03:05.520
Cedric Husler reviewed roughly 500 pages where Chat GPT had supplied at least 5% of traffic.
00:03:05.599 --> 00:03:12.000
He reported about 31,000 fewer ChatGPT attributed visits per day and about 26,000 more direct visits on those same pages.
00:03:12.159 --> 00:03:16.479
He also found that direct rows on two out of three pages that lost ChatGPT traffic.
00:03:16.639 --> 00:03:19.120
That is strong evidence of an attribution break in his sample.
00:03:19.199 --> 00:03:21.919
It is not a census of
Load full transcript
every ChatGPT account or every site.
00:03:22.080 --> 00:03:27.360
OpenAI also did not publish an explanation in the sources we reviewed, so we should not claim that the change was intentional.
00:03:27.520 --> 00:03:30.479
Profound tracked links inside its collected ChatGPT answers.
00:03:30.560 --> 00:03:33.520
It saw the share with tag source links fall sharply, then recover.
00:03:33.680 --> 00:03:39.919
By the morning of October 16th, approximately 99% of the cited answers it collected again included at least one tag source link.
00:03:40.080 --> 00:03:41.520
That recovery helps current measurement.
00:03:41.599 --> 00:03:43.199
It does not repair the historical reports.
00:03:43.360 --> 00:03:49.199
If the original visit arrived without a tag or referrer, changing a channel rule today cannot recreate information that was never collected.
00:03:49.360 --> 00:03:50.479
Annotate the affected dates.
00:03:50.639 --> 00:03:54.479
Compare direct traffic on the same landing pages that used to receive ChatGPT referrals.
00:03:54.560 --> 00:03:57.280
Look at conversions, engagement, new user rates, and device mix.
00:03:57.439 --> 00:03:59.280
Test current links on desktop and mobile.
00:03:59.360 --> 00:04:03.439
Do not tell a client that AI traffic fell until the page level and outcome data support it.
00:04:03.520 --> 00:04:06.639
The third story is Google Ads channel prioritization and budget allocation.
00:04:06.800 --> 00:04:11.039
Google is rolling out optional controls across search, shopping, performance max, and demand gen.
00:04:11.280 --> 00:04:14.960
Advertisers can classify supported channels or channel groups as high or low priority.
00:04:15.120 --> 00:04:20.720
Google will try to fund the high priority group first within the campaign's target, then send the remaining budget to lower priority inventory.
00:04:20.879 --> 00:04:25.360
The second control lets an advertiser set a maximum percentage of daily budget for a low priority channel.
00:04:25.519 --> 00:04:29.199
This is more flexible than either accepting every network or excluding one completely.
00:04:29.360 --> 00:04:30.800
There are two limits to understand.
00:04:30.959 --> 00:04:34.639
First, Google calls the allocation best effort, it is not a guaranteed hard cap.
00:04:34.879 --> 00:04:40.079
Second, performance max and demand gen group search, YouTube, Maps, Gmail, and Discover Together as Google channels.
00:04:40.240 --> 00:04:43.839
The setting does not let us give each one a separate priority inside those campaign types.
00:04:44.000 --> 00:04:46.399
Anu Adegbola's reporting gets to the useful point.
00:04:46.560 --> 00:04:51.519
This gives advertisers more say over where money goes, but the control still needs verification in the actual account.
00:04:51.680 --> 00:04:53.600
Map the available groups before changing anything.
00:04:53.759 --> 00:04:55.839
Start with a small, lower priority allocation.
00:04:55.920 --> 00:05:00.959
Add a spend alert outside the platform, then judge the test by qualified leads, sales, and cost per outcome.
00:05:01.120 --> 00:05:07.279
If a setting controls spend but the extra inventory produces weak business results, the setting did its job, and the test gave us an answer.
00:05:07.360 --> 00:05:10.160
The fourth story is advertising inside personal intelligence.
00:05:10.319 --> 00:05:15.120
Google has started testing ads in AI mode for people who connect content apps through personal intelligence.
00:05:15.279 --> 00:05:17.920
Earlier this year, that connected app experience was ad-free.
00:05:18.000 --> 00:05:18.720
The test changes that.
00:05:18.879 --> 00:05:20.879
Google's help page draws an important distinction.
00:05:21.040 --> 00:05:25.120
It says ads are not based directly on a person's Gmail inbox or Google Photos library.
00:05:25.439 --> 00:05:32.399
Search may use summaries, excerpts, and inferences from connected sources to personalize the experience, and that same information may help personalize ads.
00:05:32.560 --> 00:05:34.959
Advertisers do not receive access to the connected apps.
00:05:35.120 --> 00:05:41.439
Google also says it does not use sensitive categories such as race, religion, sexual orientation, or health to personalize ads.
00:05:41.600 --> 00:05:44.480
The user chooses whether to connect the apps and can disconnect them.
00:05:44.639 --> 00:05:47.600
For advertisers, there is nothing new to target in the reviewed sources.
00:05:47.759 --> 00:05:52.480
Google did not announce an inbox audience, a photos audience, a new campaign control, or a dedicated report.
00:05:52.560 --> 00:05:54.160
The immediate change is on the user side.
00:05:54.240 --> 00:05:59.120
Some personalized AI answers can now include ads, chosen with more context than the visible query alone.
00:05:59.279 --> 00:06:00.959
Do not describe this as Gmail targeting.
00:06:01.040 --> 00:06:02.800
Watch for the placement in eligible accounts.
00:06:02.959 --> 00:06:07.759
If reporting becomes available, separate the query, the surrounding answer, the ad, and the final lead quality.
00:06:07.839 --> 00:06:12.240
The context may be richer, but the business still needs a useful offer and a landing page that finishes the job.
00:06:12.399 --> 00:06:14.560
Meta is the fifth story with ad superburst.
00:06:14.720 --> 00:06:17.759
The product lets an advertiser reserve a window from 3 to 24 hours.
00:06:17.920 --> 00:06:24.560
Meta says reach planner will estimate the price and reach, and the advertiser can request category exclusivity so a competitor cannot reserve the same window.
00:06:24.800 --> 00:06:28.720
Ad Exchanger reports that Meta commits to delivery within 5% of the forecast.
00:06:28.879 --> 00:06:30.959
This looks a lot like reservation style media buying.
00:06:31.040 --> 00:06:32.399
Mark Ritson's argument is blunt.
00:06:32.560 --> 00:06:36.959
Concentrated exposure, first position, and a large shared moment are old brand building principles.
00:06:37.120 -->
00:06:39.519
Meta has adapted them to its own inventory and buying tools.
00:06:39.680 --> 00:06:46.720
Meta says early Superburst advertisers reached a median of 34 million people per flight and saw a median 2.9 point lift in ad recall.
00:06:46.800 --> 00:06:47.839
Those are Meta's figures.
00:06:47.920 --> 00:06:52.079
The announcement does not provide pricing, a full study design, or promise that the format will generate sales.
00:06:52.240 --> 00:06:55.120
The product can make sense when a particular day has real business value.
00:06:55.279 --> 00:06:59.920
Think about a major opening, ticket release, product launch, or event where attention inside a short window matters.
00:07:00.079 --> 00:07:05.600
It is a weaker fit when customers research for weeks, the main goal is qualified leads, or the business cannot handle a sudden surge.
00:07:05.680 --> 00:07:09.439
Before buying the burst, confirm the offer, creative inventory, staffing, and follow-up.
00:07:09.680 --> 00:07:12.800
Compare the premium with ordinary delivery using the same audience and objective.
00:07:12.959 --> 00:07:14.560
Measure what happened after the reach window.
00:07:14.720 --> 00:07:16.560
Guaranteed delivery is not guaranteed demand.
00:07:16.639 --> 00:07:18.639
The last story is Meta's updated commercial terms.
00:07:18.800 --> 00:07:28.480
The US preview says that starting October 30th, a business authorizes Meta or an entity acting for it to use crawlers, bots, and similar software on domains associated with the business's products or services.
00:07:28.560 --> 00:07:32.240
The clause also allows Meta to create test credentials and periodically pull updates.
00:07:32.399 --> 00:07:37.279
The current US wording says the access supports advertising and business products, plus integrity and security.
00:07:37.439 --> 00:07:42.399
Meta's separate crawler documentation says Meta External Ads supports advertising and business products.
00:07:42.480 --> 00:07:46.399
It also documents Meta External Agent for foundation model training or direct indexing.
00:07:46.560 --> 00:07:49.920
The uncertainty comes from the edit history, an earlier version named training directly.
00:07:50.079 --> 00:07:56.240
Meta removed that sentence from the US page after legal experts argued that it could authorize much broader use of business and publisher websites.
00:07:56.399 --> 00:08:03.040
Search Engine Journal reported that, as of October 19th, a UK English preview still contained the older training language while the US version did not.
00:08:03.279 --> 00:08:09.120
Meta says the terms do not change how it uses crawlers, do not supersede robots' preferences, and are not about scraping news sites.
00:08:09.279 --> 00:08:12.480
The reporters and legal experts found the language broader than that explanation.
00:08:12.560 --> 00:08:14.720
The versions were still moving during the reporting window.
00:08:14.800 --> 00:08:17.279
The practical response is documentation, not panic.
00:08:17.439 --> 00:08:19.279
Save the exact terms shown to the business.
00:08:19.360 --> 00:08:22.800
List the domains connected to pages, ads, catalogs, and business suite.
00:08:22.959 --> 00:08:25.759
Review the relevant crawler rules and server logs with the web team.
00:08:25.920 --> 00:08:28.240
Bring legal counsel into the contract question when needed.
00:08:28.399 --> 00:08:32.399
A technical preference in robots.text and a contractual permission are separate controls.
00:08:32.480 --> 00:08:34.320
Keep these six lines handy for client calls.
00:08:34.480 --> 00:08:38.879
Google is putting more links into AI answers, but we still need to see whether those links send visits.
00:08:39.039 --> 00:08:42.799
We will track the pages and conversions that matter instead of counting citations alone.
00:08:42.960 --> 00:08:45.279
The ChatGPT traffic drop may be a reporting gap.
00:08:45.360 --> 00:08:49.200
We will compare direct traffic and conversions on the same landing pages before calling it a loss.
00:08:49.360 --> 00:08:53.440
Google Ads now offers more control over where budget goes, but the limit is best effort.
00:08:53.519 --> 00:08:56.639
We will use spend alerts and judge the test by qualified leads or sales.
00:08:56.799 --> 00:09:02.399
Google is testing ads and personalized AI answers, but advertisers cannot target or read somebody's Gmail or photos.
00:09:02.559 --> 00:09:05.039
We will watch the placement and measure lead quality if it appears.
00:09:05.200 --> 00:09:08.000
Meta's reserve format is built for a real launch or event.
00:09:08.080 --> 00:09:11.919
We will compare the price with ordinary delivery and measure what happens after the reach spike.
00:09:12.000 --> 00:09:15.919
Meta's new terms add website crawling rights, and the wording has varied by addition.
00:09:16.080 --> 00:09:20.960
We will document the exact version, review the connected domains, and check crawler controls before October 30th.
00:09:21.039 --> 00:09:21.600
That is the week.
00:09:21.759 --> 00:09:22.559
Keep influencing.