Google May 2026 Core Update: Recovery Plan

The short answer: Google's May 2026 core update ran from 21 May to 2 June. Google published no explanation of what it was aiming at, so anyone who tells you exactly what it rewarded is guessing. What you can do is confirm whether this update is really the cause of your drop, score your affected pages against the questions Google itself publishes, and sort each page into keep, improve, merge or remove. Recovery, where it comes, takes months and often shows only at the next core update.

This guide gives you the dates to compare, a 20-point page audit, a triage table and one page audited from start to finish.

What we know, and what we do not

The facts, from Google's own Search Status Dashboard:

Update	Started	Finished
March 2026 core update	27 March	8 April
May 2026 core update	21 May	2 June
June 2026 spam update	24 June	26 June
August 2026 spam update	18 August	20 August
September 2026 spam update	24 September	Check the dashboard

The May update landed in the same week as Google I/O, where Google announced a redesigned search box and a new default model for AI Mode. Search Engine Journal noted that Google published no companion post and no stated goals for the update.

Third-party tracking does not give a tidy story either. SISTRIX's lists of the biggest movers in the US and UK had famous names on both sides: YouTube, Facebook and Yelp gained visibility, while Reddit, Amazon and Wikipedia lost some. That is not a simple "big brands won" or "forums lost" pattern.

So we work from what Google does say, every time: a core update re-assesses the whole index, it does not target individual sites, and a page can fall simply because other pages were judged more useful.

First, confirm it was this update

Plenty of drops blamed on May had other causes. Check before you rewrite anything.

Compare the right weeks. Google's advice is to wait at least a week after a core update finishes, then compare a week after with a week before it started. In Search Console, compare 14 to 20 May with 10 to 16 June. That window sits before the June spam update, so the two do not get mixed up.
Look at when the line bends. A fall that starts on 24 June, 18 August or 24 September points to a spam update. A fall that starts on a date in none of the rows above is probably something on your side: a site migration, a tracking change, a seasonal dip.
Go page by page. Open the Pages tab, sort by clicks lost, and export the top 50. Total traffic hides the pattern. Usually a handful of pages or one section carries most of the loss.
Check position, not only clicks. If average position held and clicks fell, you were not demoted. An AI Overview is probably answering the query above you. That is a different problem, covered in our post on AI Overviews.

Google also gives a sense of scale. A slip from position 2 to 4 needs no drastic action. A slide from 4 to 29 deserves a full assessment.

Score each page out of 20

Google publishes a list of questions it wants site owners to ask about their own content. We have turned the ones that matter most into a scoring sheet. Give each line 0 (no), 1 (partly) or 2 (clearly yes).

#	Question	What a 2 looks like
1	Does it contain information that is not on the pages already ranking?	Your own data, test, photos, pricing or process
2	Is it written or reviewed by someone who clearly knows the subject?	A named author with a role and a reason to be trusted
3	Does it answer the main question in the first few lines?	A reader could stop after the first paragraph and be satisfied
4	Does it cover the topic fully, without padding?	No section exists only to add length
5	Are the facts current and checkable?	Dates, prices and rules are up to date, with sources linked
6	Does the title describe the page honestly?	No exaggeration, no bait
7	Would a reader finish without needing to search again?	The obvious follow-up questions are answered
8	Is it the only page on your site targeting this question?	No near-duplicate posts competing with it
9	Is the page pleasant to use on a phone?	Loads fast, no pop-up covering the answer
10	Would you publish it if search engines did not exist?	It serves your actual customers

How to read the score: 16 to 20 is a strong page, so leave it alone. 10 to 15 is worth improving. Under 10, ask whether the page should exist as a separate page at all.

Keep, improve, merge or remove

Use the score together with the page's Search Console data.

Decision	When it applies	What to do
Keep	Score 16 or more, small or no drop	Nothing. Do not edit a healthy page out of nervousness
Improve	Score 10 to 15, the topic matters to your business	Rewrite to fix the lines that scored 0. Add what only you can add
Merge	Two or more pages answer the same question, each weakly	Combine into the strongest URL, redirect the others to it
Remove	Score under 10, no real audience, written only to catch a keyword	Delete or noindex. Redirect if it has links or a close replacement

Deleting is the last resort. Google's own guidance says to remove content only when it cannot be salvaged. On most sites we see, merging does more good than deleting.

A worked example

Take a made-up skincare brand. The page is a blog post titled "Best sunscreen for oily skin". The numbers below are invented and rounded, to show the method.

Clicks, 14 to 20 May: 1,000. Clicks, 10 to 16 June: 400.
Average position: 4 before, 11 after. So this is a real demotion, not an AI Overview effect.

The audit:

Line	Score	Why
1. Original information	0	A list of general tips found on every skincare blog
2. Expert author	0	Published under "Admin"
3. Answer first	1	The recommendation appears after 300 words of introduction
4. Full coverage	1	Nothing on reapplication or wearing it under make-up
5. Current and checkable	1	SPF claims have no source
6. Honest title	2
7. No need to search again	1
8. Only page on this question	0	Two older posts cover "sunscreen for acne-prone skin" and "matte sunscreen"
9. Phone experience	2
10. Would publish anyway	1

Total: 9 out of 20. The topic matters to the brand, so the decision is improve plus merge:

Merge the two older posts into this URL and redirect them.
Open with a three-line answer: which type of sunscreen suits oily skin and why.
Add what only the brand has: its own wear test on a handful of staff over a working day, with dated photos, and the formulator's explanation of why gel textures behave differently.
Put the formulator's name, qualification and role on the page.
Link the SPF and reapplication advice to a dermatology source.

That takes the page to roughly 17. Whether rankings return is Google's decision, but the page is now one a reader would choose.

What not to do

Do not change the publish date without changing the page. Google lists this as a sign of search-first content.
Do not mass-delete in a panic. You will remove pages that were fine.
Do not add 500 words to everything. Google has said plainly that there is no preferred word count.
Do not use AI to produce twenty new pages to "make up" the traffic. Google lists mass-produced pages with no added value as a mark of search-first content.
Do not tinker weekly. Make the changes once, properly, and record the date.

A realistic timeline

Week 1: Run the date comparison. Export the 50 pages that lost the most clicks.
Weeks 2 and 3: Score each page. Fill in the triage table.
Weeks 4 to 8: Do the merges and redirects first, then rewrite the "improve" pages, most valuable first.
After that: Leave them. Google says improvements can take anywhere from a few days to several months to register, and some changes only show at the next core update. There have been two in 2026 so far, about eight weeks apart.

Where this fits

A page audit like this is the core of our SEO work, and it is the same work that gets a page quoted in AI answers. Google's own guidance says its AI features run on the same ranking and quality systems, which is why we plan SEO together with answer engine optimization and generative engine optimization. If the audit shows the site itself is slow or awkward on a phone, that is a website development job.