Google's Spam Update Is Still Rolling Out. I Checked 68 Local Rankings Before Blaming It.

Every ranking drop this month will get blamed on the spam update. Most of them should not be.

The facts first. Google started the September 2026 spam update on September 24, it applies to every language worldwide, and Google said the rollout may take up to two weeks. It is listed on the Search Status Dashboard. It is the fourth spam update this year, and Search Engine Journal points out the first three finished in 19 hours and 30 minutes, 2 days and 1 hour, and 2 days and 16 hours. This one gets two weeks. Google has not said what it targets.

A two-week window is a long time for a single event to be the explanation for everything. Reviews change. Competitors post. AI answers push clicks around. A business owner calls me on day nine of the rollout because a keyword fell, and the easy answer is the update. The easy answer is also the one nobody checks.

So I checked before anybody called. I took every local keyword I track that had a map scan before September 24 and another one after, and I compared the movement to the same kind of month without this update in it. Then I looked at the organic side in Search Console for the five sites I have connected.

68 keywords, 15 local businesses, two comparison windows

The test group: 68 keywords across 15 local businesses, each scanned on a geo grid once between August 25 and September 23 and again between September 25 and October 1. The median gap between scans was 25 days. The businesses are the usual local mix: HVAC, roofing, gutters, windows, plumbing, restaurants, a chiropractor, a painter, a restoration company.

Comparison one is the closest thing I have to the same experiment: 270 keyword pairs across 60 businesses where the first scan landed before the August spam update started on August 18 and the second landed after it finished, August 21 to September 23. Median gap 26 days. Comparison two is a quiet stretch with no announced update in it: 150 pairs across 36 businesses, first scan August 21 to September 12, second September 8 to 23.

Median movement in average map position. September window: 0.86 positions. August update window: 0.50. Quiet stretch: 0.46. Keywords that got worse by two or more positions: 10 percent in September, 7 percent around the August update, 5 percent in the quiet stretch. Keywords that improved by two or more: 6, 7, and 6 percent.

That looks like a lean downward until you read the rows. Six of the seven keywords that fell by two or more positions started at an average position of 11.75 or worse, with share of local voice at 1.23 percent or zero. One grid point out of 81. Those businesses were already invisible for those terms. Going from almost invisible to slightly more invisible is noise with a decimal point.

The drops that matter are the ones on keywords that were winning. Share of local voice fell by 10 points or more on 7 keywords and rose by 10 or more on 1. The big one: a roofer went from 47 percent of the grid to 1.3 percent for roof repair in its home suburb. A gutter company lost 16 points on its main service term and a window company lost 15. On the other side, a restaurant went from 15 percent to 58 percent for its cuisine near me, and its main city term went from 91 to 100.

Organic next. Search Console reports in rolling 28-day windows for the five sites I have connected, so the latest window, ending September 28, holds five post-rollout days. From the window ending September 23 to that one, average position improved on all five sites, by 0.6 to 2.1 positions. Total clicks went from 800 to 790. In the five days before the update, the same five sites also improved on all five, by 0.8 to 2.5 positions, and clicks went from 761 to 800. Same direction. Same size. The one thing that sped up: impressions fell 6.3 percent, against 2.5 percent the five days before.

The update did not visibly touch my local clients. Yet.

Map rankings for the businesses I track moved a little more than usual and leaned slightly down, but almost all of the extra movement sits in keywords nobody was winning. The organic trend line did not bend. If the September spam update hit local service businesses that follow the rules, it did not show up in 68 keywords and five Search Consoles in its first week.

That is the boring answer, and it matches how spam updates work: they tune the systems that catch sites breaking Google's spam policies. Semrush's write-up on this update says it plainly: if you follow the policies, the most likely outcome is nothing. The interesting part is what a lazy read of the same data would have said. Seven keywords fell by two or more positions. Seven lost a big chunk of the map. Put those in a client email with the words spam update in the subject and you have a story. It would be the wrong story.

The roofer is the case in point. A drop from 47 percent of the grid to about 1 percent in under a month is not an algorithm nudging a page. It is the kind of thing that happens when the listing changes, the scan center moves, a competitor opens next door, or the profile gets flagged. Those get checked first. The update gets blamed last, and only if the timeline lines up and the organic side fell too.

And note the impressions. Fewer impressions with the same clicks and better average position usually means the site stopped appearing for queries it was ranking badly for anyway. That is a measurement wrinkle, not a penalty, and it is the same reporting gap I wrote about in the hidden clicks piece.

How I triage a drop during a two-week rollout

One. Write the date down. September 24, 9:15 a.m. Pacific. Every drop gets placed on that timeline before anyone says the word spam. A drop that started on September 20 is not the update.

Two. Separate the map from the web. A map drop with flat organic rankings is almost always a listing problem: categories, reviews, proximity, a suspension, a duplicate. Check the profile before you check the algorithm. A drop in both, starting inside the window, is the one worth worrying about.

Three. Ignore the floor. If a keyword was sitting at 1 percent share of voice, its movement tells you nothing. Report the keywords that were winning and nothing else.

Four. Look at who moved up. If the same two competitors gained on every keyword you lost, that is a competitor story. If thin, templated, or parasite pages jumped, that is a spam story and it may reverse when the rollout finishes.

Five. Wait for the dashboard. Google marks the rollout complete on the status page. Re-run the comparison then, not on day three. If the pages you lost are the ones that say nothing new, the fix is the one from the sameness test, not a disavow file.

Put the timeline and the rankings on one screen

What I did above took two exports and a spreadsheet. In Semrush One, Position Tracking records your target keywords every day, so you can lay the September 24 start date over the line and see whether a drop began inside the window or before it. That one question kills most false alarms.

The other two pieces answer the other questions. Organic Rankings shows a competitor's visibility over time, so you can see whether the domains that passed you surged during the rollout. And the AI Visibility reports in Semrush One track mentions and citations across ChatGPT, Perplexity, and Google AI Mode, which tells you whether a traffic drop is a Google ranking problem or an AI answer problem. During a two-week rollout you want all three, because the update is never the only thing moving.

Blame the update last.

68 keywords and five Search Consoles is one week of one rollout, not a study, and the rollout is not finished. I will run the comparison again when Google marks it complete. But the first week says what most weeks say. The map moved about as much as it always does. The organic line kept its slope. The biggest drop has a better suspect than Google.

The spam update is real. It is also the most convenient excuse in local SEO for the next two weeks. Check the timeline, check the listing, check who passed you. Then decide.