“My Emails Are Suddenly Going to Spam”: Next 24h

Quick answer: One seed message in spam, or a falling open rate, is not yet an incident. Bounce codes (Gmail 4.7.2x, 4.7.3x, 5.7.2x or 5.7.30, Outlook 550 5.7.515), a Postmaster Tools spam rate of 0.3% or more, or invoices in spam too, are. Next 24 hours: hold cold sends on that domain only, read the bounce codes, check one message’s headers. Gmail restores mitigation eligibility after 7 consecutive days below 0.3%.

My emails are suddenly going to spam: what is actually happening

When mail starts landing in spam within days of switching on an AI email agent or a new outbound sequence, one of three things has usually changed, and each leaves a different trace.

• Authentication broke. The new tool sends on your behalf but signs with its own DKIM domain, or its SPF include pushed your record past Microsoft’s warning point of 10 DNS lookups. Gmail requires bulk senders to align the From: header domain with either the SPF domain or the DKIM domain. This failure is usually loud: it tends to produce bounce codes within hours, although Google’s enforcement table allows for “spam foldering” instead of a code, so a missing code does not clear it.
• Complaints rose. Recipients who did not expect your message clicked “Report spam”. This failure is slow and quiet: Gmail’s user-reported spam rate moves first, placement follows, and the dashboards that show it run about a day behind.
• A blocklist listed you. Spamhaus and others list IPs and domains that hit spam traps or send at suspicious volumes. Spamhaus’s own warm-up guidance says a new or dormant IP should not send more than 1,000 emails a day in its first few days, and that failing to warm up “may result in the IP or domain being listed”.

AI outbound makes all three more likely on the same day, because it changes the sending tool, the volume and the audience at once. Google’s sender guidelines are specific about the volume part: “immediately doubling previously sent volumes suddenly could result in rate limiting or reputation drops.” The requirements themselves are covered in the 2026 Gmail, Outlook and Yahoo bulk-sender rules; this page is about the day after they bite.

Bounce codes name the cause of a spam-folder problem; without them, check one message’s headers before you blame reputation.

Is it as urgent as it feels?

Often not. Google’s guidelines say plainly that “some legitimate messages may be marked as spam”, and one test message landing in your own junk folder is a single data point. Nor is a falling open rate proof of anything: Google states that it “doesn’t track open rates” and that “low open rates aren’t necessarily an accurate indicator of deliverability or spam classification issues.”

It is urgent when any of these is true:

• Your sending tool shows bounces with a named code: Gmail’s 4.7.2x or 4.7.3x rate-limit codes, its 5.7.2x or 5.7.30 block codes, or Outlook.com’s “550; 5.7.515 Access denied”.
• Google Postmaster Tools shows a user-reported spam rate at or above 0.3%. Google says rates above 0.1% already harm inbox delivery for bulk senders, and a bulk sender above 0.3% is ineligible for mitigation until the rate stays below 0.3% for 7 consecutive days.
• Mail you did not change is also going to spam: invoices, password resets, staff replies. That usually means the reputation damage is on the domain itself, not on one campaign.
• Your domain or sending IP appears on a Spamhaus listing.

If none of those is true, you have a placement wobble, not an incident. Keep sending at your current volume, not more, and look again tomorrow when the dashboards have caught up.

The dashboards cover consumer inboxes, not your prospects’ company inboxes

This matters most for B2B outbound. Google’s sender guidelines FAQ says: “The Email sender guidelines don’t apply to messages sent to Google Workspace accounts. Sender requirements and Google enforcement apply only when sending email to personal Gmail accounts.” The Postmaster Tools dashboards likewise show data about email you send “to personal Gmail accounts”. Microsoft’s high-volume sender requirements are scoped to Outlook.com, which the announcement calls “our consumer service”: hotmail.com, live.com and outlook.com addresses.

If your AI agent emails procurement managers at company domains, those addresses are not personal Gmail or Outlook.com accounts, so neither consumer dashboard describes what happened to them, even when the company’s mail is hosted by Google or Microsoft. On a low-volume day Postmaster may show nothing at all: Google says dashboards “might not include all data on days when your outgoing email volume is low.” That does not mean company mail filters ignore your reputation. It means the two free dashboards cannot tell you, so your own logs have to.

Evidence source	Which recipients it covers	What you need to use it	How fresh it is
Your sending tool’s bounce and deferral log	Every recipient you sent to	Access to the raw SMTP responses, not just a “bounced” flag	Immediate
Headers of a message delivered to a seed mailbox you control	That one mailbox provider	A Gmail, Outlook.com and company-domain seed address	Immediate
Google Postmaster Tools	Personal Gmail accounts only	A verified domain; some dashboards count only DKIM-authenticated mail	Typically updated within 24 hours, can take longer; Compliance status can take up to 7 days to reflect a fix
Microsoft SNDS	Outlook.com consumer mailboxes	Proof that you own the sending IP range; useless on a shared IP you do not control	Daily, by activity period
Spamhaus IP and Domain Reputation Checker	Anyone whose mail server queries Spamhaus	Your sending IP and domain	Current listing state

What to do in the next 24 hours

Work in this order. The early steps cost nothing and need no vendor.

Hour 0 to 1: hold the cold sequence on the affected domain only

Stop new cold sends from the affected domain for the rest of the day. Leave transactional mail, support replies and other domains alone. Do not change DNS records yet and do not buy a new domain yet; you do not know the cause.

This is a triage pause, not a strategy. Google’s own instruction for bounces and deferrals is to “reduce the sending volume until the SMTP error rate decreases. Then, increase slowly again”, and it also says to “send email at a consistent rate” and avoid bursts. A week-long stop followed by a full-volume restart is a burst. Plan to resume at reduced volume tomorrow or the day after.

Hour 1 to 3: read the bounce codes, split by mail host

Export yesterday’s sends with the full SMTP response for every failure. A code names its cause far more precisely than any dashboard. Then split the results by the recipient’s mail host, because a Gmail fix and a Microsoft fix are different jobs. This short script groups a send log by the recipient domain’s MX record:

import csv, sys, subprocess, collections
def host(domain):
    mx = subprocess.run(['dig', '+short', 'MX', domain], capture_output=True, text=True).stdout.lower()
    if 'google.com' in mx or 'googlemail.com' in mx: return 'Google (Gmail/Workspace)'
    if 'outlook.com' in mx: return 'Microsoft (Outlook.com/365)'
    return 'Other/none'
cache, tally = {}, collections.Counter()
for row in csv.DictReader(open(sys.argv[1])):
    d = row['recipient'].rsplit('@', 1)[1].lower()
    cache.setdefault(d, host(d))
    tally[(cache[d], 'bounced' if row['result'].startswith('bounced') else 'delivered')] += 1
for (h, r), n in sorted(tally.items()): print(f"{h:28} {r:9} {n}")

We ran it on 30 September 2026 against a six-row test file (recipients at gmail.com, outlook.com, microsoft.com, google.com and example.com, with two rows marked as bounced) using live MX lookups. Output:

Google (Gmail/Workspace)     bounced   1
Google (Gmail/Workspace)     delivered 2
Microsoft (Outlook.com/365)  bounced   1
Microsoft (Outlook.com/365)  delivered 1
Other/none                   delivered 1

Your file needs two columns, recipient and result. Failures concentrated on one host are the finding.

Hour 3 to 6: check one real message’s headers and your DNS

Send one message from the AI tool to seed mailboxes you control, open “Show original” (Gmail) or the message source (Outlook), and save it as a .eml file. This script reports what the receiving server concluded, which DKIM keys signed the message (a tool can add its own signature alongside yours, so it prints every one), and whether one-click unsubscribe meets RFC 8058, which requires an HTTPS URI in List-Unsubscribe, a List-Unsubscribe-Post header reading “List-Unsubscribe=One-Click”, and a DKIM signature whose h= tag covers both headers:

import sys, re, email
from email import policy
msg = email.message_from_binary_file(open(sys.argv[1], 'rb'), policy=policy.compat32)
ar = ' '.join(str(h) for h in msg.get_all('Authentication-Results', []))
ar = re.sub(r'\s+', ' ', ar)
for mech in ('spf', 'dkim', 'dmarc'):
    m = re.search(mech + r'=(\w+)', ar)
    print(f"{mech:6} {m.group(1) if m else 'MISSING'}")
covered = False
for raw in msg.get_all('DKIM-Signature', []):
    sig = re.sub(r'\s+', '', str(raw))
    tags = dict(kv.split('=', 1) for kv in sig.split(';') if '=' in kv)
    signed = tags.get('h', '').lower().split(':')
    covered = covered or ('list-unsubscribe' in signed and 'list-unsubscribe-post' in signed)
    print(f"dkim d= {tags.get('d','?')}  s= {tags.get('s','?')}  -> dig +short TXT {tags.get('s','?')}._domainkey.{tags.get('d','?')}")
lu, lup = msg.get('List-Unsubscribe', ''), msg.get('List-Unsubscribe-Post', '')
ok = ('https://' in lu and lup.strip() == 'List-Unsubscribe=One-Click' and covered)
print('one-click (RFC 8058):', 'OK' if ok else 'NOT MET')

We ran it against two synthetic test messages we built for this page. The first imitates a common AI-outbound failure: the tool signs with its own domain (outreach.example.org) while the From: address is on example.com, and the only unsubscribe option is a mailto link:

spf    pass
dkim   pass
dmarc  fail
dkim d= outreach.example.org  s= s1  -> dig +short TXT s1._domainkey.outreach.example.org
one-click (RFC 8058): NOT MET

SPF and DKIM both “pass”, and DMARC still fails, because neither passing domain matches the From: domain. The second message, signed by example.com with both unsubscribe headers inside the signature, returned dmarc pass and one-click (RFC 8058): OK. The script checks that the headers are present and listed in the DKIM h= tag; it does not verify the signature cryptographically or test that your unsubscribe endpoint works, and RFC 8058 adds that the sender “MUST NOT return an HTTPS redirect” to the POST.

Then confirm the published records. Run against gmail.com and example.com on 30 September 2026:

$ dig +short TXT gmail.com | grep -i 'v=spf1'
"v=spf1 redirect=_spf.google.com"
$ dig +short TXT _dmarc.gmail.com
"v=DMARC1; p=none; sp=quarantine; rua=mailto:mailauth-reports@google.com"
$ dig +short TXT _dmarc.example.com
"v=DMARC1;p=reject;sp=reject;adkim=s;aspf=s"

Substitute your own domain, and use the selector the script printed for the DKIM lookup. If no d= line shows your domain (or a subdomain of it), you are in the 4.7.32 row of the table below; if the key lookup returns nothing, the signature cannot verify.

Hour 6 to 24: read the dashboards, check the blocklists

Open Postmaster Tools’ Compliance status, Spam rate and Authentication dashboards, remembering the data describes yesterday at best. If you run your own sending IPs, open SNDS. Look up your sending IP and domain in the Spamhaus checker. Then match what you found against the table below.

Do not move DMARC to p=reject tonight. Gmail’s bulk-sender minimum is a DMARC record with a policy of none, and gmail.com itself published p=none when we queried it on 30 September 2026. Microsoft does say that “p=reject is the most effective at thwarting domain spoofing”, but in the same answer advises “moving gradually (none → quarantine → reject) to avoid unintended mail loss.” If any legitimate stream is still misaligned, tightening policy mid-incident turns its spam-folder placements into quarantines or rejections. The current DMARC specification, RFC 9989 (May 2026), replaces RFC 7489 and lists the old pct rollout tag among its removed tags, so a guide telling you to set pct=10 predates it.

The bounce-code-first triage table

Rows with a code are the most specific evidence you will get, so read top to bottom.

What you see	Where you see it	Most likely cause	First fix	Urgency
4.7.27 (rate limited) or 5.7.27 (blocked): SPF did not pass	Bounce log, Gmail recipients	The AI tool’s servers are not in your SPF record, or the record fails	Add the tool’s include; keep SPF under 10 DNS lookups	Today
4.7.30 or 5.7.30: DKIM did not pass	Bounce log, Gmail	The tool’s DKIM key is not published on your domain	Publish the tool’s DKIM key on your domain (custom DKIM)	Today
5.7.26: message not authenticated	Bounce log, Gmail	Neither SPF nor DKIM passing for the tool’s traffic	Set up both; Google requires both for bulk senders	Today
4.7.31: no DMARC record or no policy	Bounce log, Gmail	No DMARC record on the sending domain	Publish p=none with an rua reporting address	Today
4.7.32: From: not aligned with SPF or DKIM domain	Bounce log, Gmail	SPF and DKIM pass, but for the tool’s domain	Sign with your own domain so DMARC aligns	Today
4.7.23 or 5.7.25: PTR record missing or mismatched; 4.7.29 or 5.7.29: no TLS	Bounce log, Gmail	Self-hosted or misconfigured sending server	Fix reverse DNS and TLS, or move to a provider that has both	Today
4.7.28: rate limit exceeded	Bounce log, Gmail	Too much volume too fast for the domain or IP	Stop at least 10 minutes, then one connection, adding connections one at a time	Today
550; 5.7.515 Access denied	Bounce log, Outlook.com, Hotmail, Live recipients	Domain sending over 5,000 a day fails SPF, DKIM or DMARC	SPF must pass, DKIM must pass, and DMARC must be at least p=none, aligned with SPF or DKIM	Today
Spam rate at or above 0.3%	Postmaster Tools, Spam rate	Recipients reporting your mail	Hold cold sends on the domain; mitigation needs 7 consecutive days below 0.3%	Today
Spam rate between 0.1% and 0.3%	Postmaster Tools, Spam rate	Targeting or message problem starting to bite	Cut volume to recently engaged recipients; review who the agent is emailing	This week
High spam rate suddenly drops to 0, or disappears	Postmaster Tools, Spam rate	Gmail now filters many of your messages to spam, so fewer recipients see them to report	Treat as worse, not better	Today
Red status: spam verdict on more than 90% of verdicts	Microsoft SNDS, your own IPs only	IP reputation at Outlook.com	Check complaints via JMRP; go to the Outlook.com postmaster support route	Today
Listed in Spamhaus CSS, DBL or SBL	Spamhaus checker	Spam traps, volume spike, or a compromised system	Fix the cause first; removal routes differ by list (next section)	Today
Invoices and password resets also going to spam	Customer complaints, your own seed tests	Cold outbound damaged the reputation your core mail relies on	Stop cold sends from that domain; apply the invoice-domain rule	Today
Open rate fell; no bounces; dashboards normal	Your email tool’s reports	Not evidence on its own; Google does not track open rates	Seed test, then recheck Postmaster tomorrow	Not yet
One seed message in spam	Your own inbox	Normal filtering noise	Repeat across several seeds and providers	Not yet

Two rows need a warning. Google’s 4.7.28 guidance adds that if the error does not say whether the DKIM, SPF or IP quota was exceeded, “assume that all three are affected”, and if the single connection fails, wait another 10 minutes; the sending ceilings behind that code are set out in Gmail and Outlook sending limits for AI email agents. And Microsoft describes the 5.7.515 outcome inconsistently: its April 2025 announcement carries an update saying failing messages will be rejected with that code, while the paragraph beneath it, and the Outlook.com postmaster site read on 30 September 2026, still describe routing to Junk first. Expect either.

The next 7 days: resume slowly and let the data catch up

1. Resume at reduced volume, to your most recently engaged recipients first. Google advises starting “with a low sending volume to engaged users” and increasing slowly, at a consistent daily rate. Spamhaus’s warm-up advice for a new or dormant IP adds: avoid doubling the daily volume in the first week.
2. Expect the numbers to lag the fix. Google says “it can take time for improvements in spam rate to reflect positively on spam classification”, that after acting on a high spam rate you should “wait up to 7 days” for it to fall within compliance, and that a fixed Compliance status item can take up to 7 days to show. Mitigation eligibility returns after 7 consecutive days below 0.3%.
3. Handle blocklists by their own rules. Spamhaus’s CSS listings “generally expire three (3) days after the last detection” and re-list on re-detection. Most DBL domain listings expire automatically once the activity stops; an approved removal is processed immediately, with some systems lagging up to 24 hours, and the form “does not guarantee removal”. SBL removals must be requested by the ISP responsible for the IP. Spamhaus says there is never a fee, and that any offer to remove a listing for a fee is a scam.
4. At Outlook.com, give a new IP time. Microsoft says a new IP “can expect to be fully ramped within a couple of weeks or sooner”, provided complaint rates stay low.
5. Stop verifying addresses by probing mail servers. Microsoft blocks what it calls namespace mining, “verifying email addresses without sending (or attempting to send) emails to those addresses”. Some list-cleaning tools work that way; check how yours does.

A bulk sender over Gmail’s 0.3% spam-rate line stays ineligible for mitigation until it has spent 7 consecutive days below it.

The fix that stops it recurring: the invoice-domain rule

The invoice-domain rule: never send cold AI outbound from the domain your invoices, password resets and staff email come from. If a cold sequence goes wrong, the damage should land on a stream you can afford to pause, not on the mail your customers pay you through.

The mailbox providers point the same way. Yahoo’s sender requirements say “Don’t send bulk/marketing email from the same IPs you use to send user mail, transactional mail, alerts, etc.”, because “each IP and DKIM domain has a reputation”. Google’s guidelines are more cautious: ideally send everything from one IP address, but if you must use several, use a different one for each message type, and keep messages of the same category on the same From: address.

The rule has limits, and they are worth knowing before you build on it:

• A subdomain does not fully separate you at Google. Google counts every message from the same primary domain toward its 5,000-a-day bulk-sender threshold, and its Compliance status dashboard reports only at primary-domain level. A subdomain separates the From: address and DKIM signing, not Google’s bulk-sender status.
• Separation protects your invoices; it does not make cold mail welcome. Spamhaus’s position is that “no email should be sent until and unless there is direct and verifiable permission”. Cold AI outbound does not meet that standard, whatever domain it comes from, and recipients who did not expect it are the ones who report it.

The other recurrence fixes are unglamorous. Put RFC 8058 one-click unsubscribe headers on every outbound message the agent sends, and remove people within 48 hours, which is Google’s window; Yahoo asks for 2 days. Check the legal basis for each list before an agent touches it: in Australia that means express consent or inferred consent under the Spam Act, which the sender has to be able to prove. And cap volume below what the tool allows, because more sequences to a broader list is the pattern described at the top of this page; the mechanics are in why reply rates fall as AI outbound volume rises.

Doing this by hand means running the 24-hour sequence above for every incident, plus a standing weekly check of bounce codes, seed placement, dashboards and blocklists for every sending domain. The time grows with the number of domains and tools, not with the number of emails sent. Zian, which has run outbound acquisition since 2017, treats email as one channel among phone, SMS and WhatsApp rather than the only route to a lead, which is the other way to take pressure off a single sending domain.

Frequently asked questions

Why are my emails suddenly going to spam when nothing changed?

Usually something did change: a new AI sending tool that signs mail with its own DKIM domain, a jump in daily volume, or a broader list. Google’s email sender guidelines warn that immediately doubling previously sent volumes suddenly could result in rate limiting or reputation drops. Check your bounce codes first, then the headers of one delivered message, before assuming your reputation is the problem.

How long does it take to recover from a high Gmail spam rate?

At least a week once the cause is fixed. Google’s email sender guidelines FAQ says bulk senders with a user-reported spam rate above 0.3% are ineligible for mitigation, and become eligible again when the rate stays below 0.3% for 7 consecutive days. Spam rate is calculated daily, Google also says it can take time for a lower spam rate to improve spam classification, and it recommends staying below 0.1%.

Which Gmail requirements cause rejections, and which only remove delivery support?

Google’s enforcement table lists nine issues. Five can bring temporary or permanent failure codes, or spam foldering: a From: header not aligned with authentication, mail not authenticated with both SPF and DKIM, missing forward and reverse DNS, no TLS, and messages that do not follow RFC 5322 format. Four make delivery support or mitigations unavailable: a spam rate above 0.3%, a missing DMARC record (minimum policy of none), marketing mail without one-click unsubscribe, and unsubscribe requests not honoured within 48 hours. Google’s error-code list separately includes 4.7.31, a temporary rate limit when the sending domain has no DMARC record or no DMARC policy.

Why are my Outlook emails suddenly going to spam?

For Outlook.com, Hotmail and Live addresses, check authentication first. Microsoft’s high-volume sender requirements apply to domains sending more than 5,000 emails a day and require SPF and DKIM to pass and DMARC of at least p=none, aligned with one of them. Its pages describe failing mail being sent to Junk and also being rejected with 550 5.7.515, so expect either. SNDS only helps if you own the sending IPs.

Does Google Postmaster Tools show emails sent to Google Workspace addresses?

No. Google’s Postmaster Tools dashboards page describes the data as outgoing email you send to personal Gmail accounts, and the sender guidelines FAQ says the requirements and enforcement apply only to personal Gmail accounts, not to Workspace recipients. For B2B lists, your own bounce log and seed-message headers are the evidence that covers company inboxes.

Should we move our AI outbound to a new domain?

Move it to separate cold mail from your invoices and staff email, not to escape complaints. A subdomain still counts toward the primary domain’s 5,000-a-day Gmail bulk-sender threshold, and Google’s sender guidelines FAQ says enforcement for new domains, meaning any domain that has not sent more than 5,000 emails a day to personal Gmail accounts since 1 January 2024, runs on an accelerated timetable.

Can we pay someone to get us off a blocklist?

No, and you should not try. Spamhaus says there is never any charge for removing a Spamhaus listing, that any offer to remove one for a fee is a scam, and that no third party can influence or expedite removals (Spamhaus Blocklist FAQ). Fix the cause, then use the removal route for that list: CSS listings generally expire three days after the last detection.

Where every figure on this page comes from

Figure	Who published it	Link	Date read
Keep spam rate below 0.1%; never reach 0.3%	Google (Gmail Help, Email sender guidelines and FAQ)	support.google.com	30 September 2026
Above 0.3%: ineligible for mitigation; eligible after 7 consecutive days below 0.3%; spam rate calculated daily	Google (Email sender guidelines FAQ)	support.google.com	30 September 2026
Bulk sender: close to 5,000 messages or more a day to personal Gmail; subdomains count toward the primary domain; new-domain definition (since 1 January 2024)	Google (Email sender guidelines FAQ)	support.google.com	30 September 2026
Nine-row enforcement table; unsubscribe within 48 hours	Google (Email sender guidelines FAQ)	support.google.com	30 September 2026
Error codes 4.7.23, 4.7.27, 4.7.29, 4.7.30, 4.7.31, 4.7.32, 5.7.25, 5.7.27, 5.7.29, 5.7.30	Google (Email sender guidelines FAQ)	support.google.com	30 September 2026
5.7.26 for unauthenticated mail; 4.7.28 and the 10-minute wait; doubling volume warning; open rates not tracked	Google (Gmail Help, Email sender guidelines)	support.google.com	30 September 2026
Dashboard data typically updated within 24 hours; Compliance status up to 7 days; wait up to 7 days after spam-rate fixes	Google (Gmail Help, Postmaster Tools dashboards)	support.google.com	30 September 2026
More than 5,000 emails a day; 550; 5.7.515; SPF over 10 DNS lookups may fail	Microsoft (Tech Community, Outlook high-volume sender requirements, updated 30 April 2025)	techcommunity.microsoft.com	30 September 2026
Junk-first wording for non-compliant high-volume mail	Microsoft (Outlook.com Postmaster)	sendersupport.olc.protection.outlook.com	30 September 2026
SNDS colours: green under 10%, yellow 10% to 90%, red over 90% spam verdicts	Microsoft (SNDS FAQ)	substrate.office.com	30 September 2026
New IP fully ramped within a couple of weeks; namespace mining	Microsoft (Outlook.com Postmaster, Troubleshooting)	substrate.office.com	30 September 2026
Unsubscribe within 2 days; separate bulk mail from transactional IPs	Yahoo (Sender Hub, Sender Requirements and Recommendations)	senders.yahooinc.com	30 September 2026
New or dormant IP: no more than 1,000 emails a day in the first few days; no doubling in the first week	Spamhaus (FAQ, marketing email)	www.spamhaus.org	30 September 2026
CSS listings generally expire 3 days after last detection	Spamhaus (FAQ, Combined Spam Sources)	www.spamhaus.org	30 September 2026
DBL removal processed immediately once approved; local lag up to 24 hours	Spamhaus (FAQ, Domain Blocklist)	www.spamhaus.org	30 September 2026
No fee for any Spamhaus removal; SBL removal requested by the ISP	Spamhaus (FAQ, Spamhaus Blocklist)	www.spamhaus.org	30 September 2026
One-click unsubscribe: HTTPS URI, List-Unsubscribe=One-Click, both headers DKIM-signed	IETF (RFC 8058, January 2017)	www.rfc-editor.org	30 September 2026
DMARC spec RFC 9989 (May 2026) obsoletes RFC 7489; pct tag removed	IETF (RFC 9989)	www.rfc-editor.org	30 September 2026
gmail.com DMARC record p=none; sp=quarantine	Google’s published DNS record, queried by us (dig +short TXT _dmarc.gmail.com)	DNS lookup (no web page)	30 September 2026
Outbound acquisition since 2017	Zian AI (first-party claim released by the owner)	First-party; no external page	30 September 2026