Charleston SEO Recap: More Links, No Numbers (Oct 12-18)
Google added link slots to the top of AI answers on Thursday and again refused to report clicks. Friday brought a ranking spike nobody can independently verify.
Google spent this week building more places for your website to appear inside an AI answer. On Thursday it announced four of them at once. Then a reporter asked the obvious follow-up question, which is whether anyone actually clicks those links, and Google said no.
That is the whole week in one sentence: more doors into your site, zero instrumentation on any of them. And it landed in the same month Google started actively breaking the independent tools you would otherwise use to check its work.
Five things, ordered by what they cost a Charleston business that lives off the phone ringing.
First, Google put links at the top of the AI answer
On Thursday, Hema Budaraju, a VP of product management at Google, published a post on the company blog announcing five changes to how AI Mode and AI Overviews show links. I read the announcement itself rather than anybody's summary of it, because the exact wording matters here.
The headline change is a test. Google says it will try "a new display in Search AI experiences to prominently surface these links in a dedicated section at the top of a response, with context about why each link is useful." Google's own mockup of it, which is the image on this post, puts a block labeled "Quick results from the web" above the written answer, holding one or two links, each with a sentence explaining what you get if you click.
The other four:
• Inline links inside the body of AI answers are moving higher up the response more often • The article carousels you see for timely news are rolling out to more countries and languages • The links at the bottom of AI responses are getting image thumbnails and a more clickable design • Creator and publisher names on video results are becoming clickable, linking to their Search profiles, in AI Mode
Google also said the number of people using Preferred Sources "has grown by more than 7x since the start of the year." Preferred Sources is the setting that lets a person tell Google which sites they want to see more of. Seven times growth off an unstated base is not a number I can do much with, but the direction is real and Google keeps building features that feed off it.
Here is why I care more about that top block than anything else announced this quarter. Think about what "the official place to get it done" means for a search like repairing a slate roof in Mount Pleasant, or an emergency vet on James Island at 9pm. There is no official page for either. There are twelve businesses that could all credibly claim it, plus four directories that will claim it harder. Google is now building a slot at the very top of the answer and deciding, by itself, who belongs in it.
A month ago I wrote about an AI Overview that named three Charleston plumbers for a query with no matching web pages behind it. Same machinery, new real estate. The answer gets written either way. The question is whose link sits above it.
What I would actually do about this, this week:
• Make one page on your site unambiguously the thing you do, not a tab buried under Services. One URL, one job, a phone number in the text
• Check that your Organization schema has a sameAs array pointing at your Google Business Profile, your Facebook page, and anywhere else you genuinely exist
• Make sure your name, address and phone match character for character across your site, your GBP, and your top five citations. Google's idea of "official" is built out of corroboration
• Write a plain sentence on your homepage saying what you do and where. Not a slogan. The sentence a machine would quote
Second, Google still will not tell you if anyone clicked
Barry Schwartz pushed Google on the obvious thing: fine, more links, so show us the clicks. A Google spokesperson gave him this: "We're continuing to work with website owners to understand what insights will be most helpful to inform their strategies, and we'll introduce additional metrics over time."
Schwartz's response was that click data is the number one metric in Google Ads, and that publishers have been asking for exactly this and nothing else. He is right and I will go further than he did.
Google does not need to research what insight would be most helpful. It has known since 2000. It built an entire advertising business on reporting clicks to the quarter of a cent, in near real time, to anyone with a credit card. The capability is not the obstacle. Reporting AI Overview clicks would produce a number that Google would then own publicly, forever, every quarter, and everyone who has measured it from the outside says that number is not flattering.
At this point a lot of people are thinking: "so what, I never looked at that report anyway." Fair. But this is the week the asymmetry got formal. Google is now telling you, in a blog post, that it is sending more traffic than before. It is telling you this with no figure attached, in a year when it has also been degrading your ability to measure independently. You are being asked to take growth on faith from the only party holding the data.
So build your own instrument, because nobody is handing you one:
• Turn on call tracking, or at minimum read your Google Business Profile call history every Monday • Watch direct and branded organic traffic as a pair. AI answers tend to convert into someone typing your name, not clicking your link • Add "how did you hear about us" to your intake form and actually read the answers once a month • Log your total booked jobs weekly. It is a worse metric in every way except the one that matters, which is that Google cannot touch it
Third, Friday's volatility spike, and why I cannot tell you if it was real
Friday morning, Barry Schwartz reported a sharp spike in ranking volatility and community chatter. Wincher, Semrush, Mozcast, Sistrix, Algoroo, AccuRanker, Mangools, Serpstat, DataForSEO and others all lit up at once. WebmasterWorld filled with people reporting traffic collapse, one of them writing that the morning "has been such a disaster that I'm wondering why we're even bothering to write anymore."
Google has confirmed nothing. The Search Status Dashboard still shows the September 2026 spam update, which ended October 8, as the most recent ranking entry. Barry's own algorithm-update category has nothing newer than Friday's unconfirmed post.
Now the part that makes this genuinely hard. Last week I covered the two vendor investigations showing Google is feeding deliberately altered results to rank-tracking tools, at a rate running roughly 5 to 11 percent of collected pages per day. Those are the same tools whose charts spiked on Friday.
I am not saying Friday was fake. Real humans with real analytics saw real traffic drops, and that is not tracker data. I am saying the industry instrument that normally tells us "yes, something shifted, here is how big" currently has a known contamination rate and no way to flag which rows are bad. When the thermometer is being tampered with, a reading of 103 is information, but it is not a diagnosis.
My stance: do not make a single decision based on a volatility chart this month. Decide off your own Search Console clicks, your own call volume, and your own manual spot checks of your own money terms. If your Charleston numbers held steady through Friday, you were not in it, whatever the charts looked like.
Fourth, Google quietly rewrote the instant indexing docs
Also Friday: Google rewrote three pages of its Indexing API documentation. It removed the approval form, the approval-process explanation, the initial-testing quota language, and a sitemap recommendation. Google told Schwartz it updated the page "to document the current quota limits for the Indexing API and remove an outdated form." The current limits still sit there in black and white: 200 publish requests per day, 180 metadata requests per minute, 380 requests per minute across all endpoints.
A developer in the comments summed up the confusion well: "This change makes zero sense. A new service account still gets a default quota of 200."
This is deep plumbing and most owners reading this will never touch it. Here is why it is on the list anyway. The Indexing API is officially restricted to two things: pages with JobPosting markup, and livestream pages with BroadcastEvent inside a VideoObject. That restriction is still in the docs. It has never covered your services page, your blog post, or your new location page.
If somebody is selling you an "instant indexing" add-on, this is a fine week to ask them a direct question: which Google endpoint are you calling, and does my page type qualify for it? There is a legitimate answer involving IndexNow, which Bing and others support and which we ping on every post here. There is also an answer that amounts to hammering an API that was never meant for your content. The difference is worth asking about before the next invoice.
Fifth, the model writing those answers changed on Tuesday
Gemini 4 Argon went generally available on Tuesday and reached AI Mode in Search for Google's AI Pro and AI Ultra subscribers.
That is it. That is the whole item, and I am including it because of what it implies rather than what it says. The system that composes the paragraph a potential customer reads about Charleston roofers was swapped mid-week. There is no changelog for how that affects which businesses get named, no before-and-after you can pull, no Search Console filter for it. Two weeks ago I wrote about a Meta AI page taking the top organic result for a Charleston SEO query, and the lesson is the same one: the surfaces answering questions about your business keep changing underneath you, on a schedule nobody publishes.
The practical response is unglamorous. Search your own category the way a customer would, from your phone, signed out, once a week. Write down who got named. That log is worth more than any tool subscription right now, and it costs four minutes.
What to prioritize this week
Ranked, because not all of this is equally urgent:
- Pick your one official page and make it obvious. One URL per core service, your city in the text, your phone in the text, your schema pointing at your real profiles. This is the thing Google's new top-of-answer block is built to find
- Start measuring outcomes you control. Calls, booked jobs, "how did you hear about us." One week of honest numbers beats a year of rank reports you cannot verify
- Ignore the Friday volatility charts. Check your own Search Console clicks against the prior 28 days instead, and hold any site changes until you have a clean read
- Ask your indexing vendor which endpoint they call. Two minutes, one email, potentially a line item off your bill
- Run your own category search, signed out, and log who Google named. Four minutes a week, every week
None of this is exciting. All of it is the part that still works when the instruments go dark, which is where we are right now. This is how we work with our service-business clients in the Lowcountry, and if you want somebody to set that measurement up properly rather than guessing at it, that is what our Charleston SEO work is.
See you next Sunday.