Article · October 18, 2026 · 9 min read

A new Panda: what Google changed in its docs and when to expect the 2026 core update

In early October Google updated two key content documents: its helpful content guide and its guidance on generative AI. The last core update ended on June 2, spam updates keep getting closer together, and Dan Taylor’s forecasting model puts the odds of a major update within a month at 96.8%. Here’s what changed, why people compare it with the 2011 Panda update, and how to prepare your site.

Check my site before the update The two-week plan

Google updates in 2026Confirmed rollouts and the forecast window. Arcs show gaps between spam updates.

CoreSpamDiscoverforecast window

96.8%

chance of a major update within 30 daysDan Taylor’s forecasting model · source

137

days without a core updateJune 2 to October 17, 2026 · source

spam updates in 2026vs one in all of 2025 · source

days between the last two spam updatesdown from 92, then 55 · source

In short

The helpful content guide now defines main content and names four quality attributes: effort, originality, talent or skill, and accuracy. Fake authors with AI headshots and false credentials are called deception.
On October 1, 2026, Google called manual review of all AI-generated content before publishing “critical”. That includes metadata: title elements, meta descriptions, structured data and alt text.
The last core update finished on June 2. By Taylor’s count there are usually three or four a year, so the next one is overdue. His model gives 96.8% for a major update within 30 days and a window from late October to mid-November.
This time Google updated its documentation first, and practitioners expect an update to follow. With Panda in 2011 the order was reversed: the update came first, the 23 self-assessment questions about ten weeks later. The comparison is about the target: low-value content at scale.
A forecast is not an announcement. Google doesn’t pre-announce updates and nobody knows the date. Prepare for the criteria, not for a date.

What Google changed in its documentation

The changes landed in two Search Central documents. Below is the gist of what was added, in our own words. Google’s exact wording is in the originals linked at the end of this section.

Helpful, reliable contentchanged October 1, page updated October 5, 2026

+A definition of a page’s main content
+Four attributes: effort, originality, skill, accuracy
+Fabricated author profiles are deception

Using generative AI contentupdated October 1, 2026

+Manually review and fact-check all AI content before publishing
+The same for metadata: title, description, structured data, alt
+Models predict words rather than retrieve facts: “hallucinations” happen

1. Main content and four quality attributes

Main content is any part of a page that directly helps it achieve its purpose: text, a tool, a table, reviews. It is judged on four attributes: how much human effort went into it, whether it adds original information, whether it shows skill, and whether the facts are accurate. A self-audit on these attributes is in part 2 of this series.

2. Fake authors

Google asks sites not to use deceptive authorship information, citing AI-generated headshots, made-up names and false credentials. How to check your own authors is in part 3.

3. A person reviews AI content and metadata

Google calls manual fact-checking and review of all generated content before publishing critical. What’s new is metadata: titles, meta descriptions, structured data and alt text also show up in Search, so they get the same review. The review process is covered in part 2.

4. Scale without value

AI itself isn’t banned. Generating many pages primarily to manipulate search rankings rather than to help users violates the spam policies. The policy’s own example: many pages generated with AI tools without adding value for users.

Sources: Google: helpful content, Google: generative AI, Google: spam policies, Relevant Audience: authors, Relevant Audience: AI

Why people bring up Panda

On February 24, 2011, Google announced an algorithm change that affected 11.8% of U.S. queries. It lowered sites with low value for users and sites that copied content from others. The industry called it Panda. According to Dan Taylor, Panda worked as a site-level quality classifier.

In May of the same year Google published 23 self-assessment questions. Among them: would you trust this article, would you give this site your credit card details, does the article have factual errors, would you expect to see it in a printed magazine. You can still recognize these questions in today’s guidelines.

	Panda, 2011	Autumn 2026
Target	Low-value and copied pages	Mass AI generation without review, fake authors
How it was described	23 self-assessment questions, May 2011	Main content definition and four attributes, October 2026
Order	The update on February 24, the questions about ten weeks later	Documentation first, no major update yet
Scale	11.8% of U.S. queries	Unknown: the update hasn’t happened

The similarity is the target: then and now, Google describes what it considers weak content. The difference is the order: in 2011 the questions came after the update; this time the criteria came first.

Sources: Google Blog, 2011, Search Engine Watch, Dan Taylor, SEJ

When to expect the update

The calendar at the top shows the whole year. There were two core updates: March 27 – April 8 and May 21 – June 2. Spam updates came in March, June, August and September, with shrinking gaps: 92 days, 55 days, 37 days. The September spam update took 13 days and 16 hours and finished on October 8.

Dan Taylor’s model relies on two groups of data: the calendar (25 years of intervals between updates, plus seasonality) and current SERP swings together with social media chatter. The result:

Probability of a major update, Taylor’s modelThe 90-day figure is measured from October 13, 2026; the source gives no start date for the 30-day one

Within 30 days
96.8%
Within 90 daysas published: lower than the 30-day figure
90.9%
Late October – mid-November
90–96%
December 2026
≈15%

This is one author’s forecast, not a Google statement. The model is probabilistic: it is trained on update history and SERP noise, not on inside information from Google.

Others share the expectation. On October 4, Lily Ray wrote that the next major update is close and listed the same documentation changes. Glenn Gabe and Marie Haynes also expect a significant update.

This autumn, ranking swings in trackers aren’t only about updates. According to Nozzle, since mid-September Google has been serving rank trackers part of the results with the wrong sites; Google hasn’t confirmed this. See our research Poisoned Google SERPs.

Sources: Dan Taylor, SEJ, Lily Ray, SEJ: March core update, SER: May core update, LoopLoop

Whole site or individual pages?

Opinions differ. Taylor and several others warn that mass generation and fake authors may hurt the whole site rather than individual pages, as Panda did.

Google’s core update documentation says the updates are broad and don’t target specific sites or pages. For a large sitewide drop, Google suggests assessing the whole site first and then the most affected pages. According to a recap of Search Central Live in Barcelona, Gary Illyes said core updates assess content at page level, not domain level.

The practical takeaway is the same either way: start with sections that have many similar AI-generated pages. They drag down both themselves and trust in the site.

Also from Barcelona, as reported by attendees: Gary Illyes said scaled content is becoming a bigger problem than link spam. The recap’s section on Duy Nguyen’s talk says spam updates are more frequent because of the volume of new content and that Google uses AI to catch spam.

Sources: Google: core updates, Roast: Search Central Live, Dan Taylor, SEJ

A two-week plan before the rollout

Two weeks, six tasks

Find mass-generated sectionsweek 1
Similar pages published in batches, templated “per city” pages, topics where the site has no experience.
Score key pages on the four attributesweek 1
The scoring method and checklist are in part 2.
Sort out authorshipweek 1
Is there a real person behind every byline, and can that be confirmed off the site? Details in part 3.
Review AI texts and metadataweek 2
Facts, numbers, title, description, structured data, alt. Validate the markup.
Decide on weak pagesweek 2
Improve, merge or noindex. Deleting is a last resort.
Save a baselinebefore the update
Export clicks and impressions by section from Search Console to compare week over week after the update.

After an update, Google advises waiting for the rollout to finish, then another week, and comparing that week with the week before the update. For a small drop, Google advises against drastic action. If your site grew in 2026 through volume rather than quality, the next step is not more pages but a topic structure where every page has its own job.

Source: Google: core updates

FAQ

When is the next Google core update?

There is no confirmed date: Google doesn’t pre-announce updates. Dan Taylor’s forecasting model gives 96.8% for a major update within 30 days and points to late October through mid-November 2026. The last core update finished on June 2, 2026.

What is “a new Panda”?

It’s what the SEO community calls the expected update aimed at mass AI content and fake authors. It’s compared with the 2011 Panda by its target: low-value content at scale. Google has not used this name.

Has Google banned AI content?

No. What Google expects from AI content and where the line is: part 2 of the series.

What should I do if my site drops after the update?

Wait for the rollout to finish, wait another week and compare it with the week before. For a small drop, avoid drastic changes. For a large one, assess the whole site and then the most affected pages. Recovery can take several months.

Sources

Update dates are from the Google Search Status Dashboard as reported by industry media. The forecast is Dan Taylor’s model; we did not reproduce it. Statements from Search Central Live are from an attendee recap, paraphrased, not quotes. Sources checked on October 17, 2026.

How this piece was made: written by the SurfUp editorial team with AI assistance; every fact and number was checked by hand against the sources listed above.