Email preview text — the short snippet of copy that appears in the inbox beneath the subject line — has been an email marketing best practice since mobile email clients began displaying it around 2012. Most email developers know the basics: set a meaningful preview to supplement the subject line, do not let unintended HTML text show in the preview, keep it under 90-100 characters to avoid truncation. In 2026, preview text has taken on a second function that the standard guidance does not address: it is now the primary source material for the AI-generated inbox summaries that Gmail Gemini, Apple Intelligence, and Microsoft Copilot display instead of the preview text itself for users who have AI inbox features enabled.
This creates a situation where the preview text line simultaneously serves a human reader (who sees it in the inbox preview) and an AI system (which reads it as the starting point for an inbox summary). The optimal preview text for both audiences is the same text — but the characteristics that make preview text effective for human readers and the characteristics that make it useful for AI summarisation are not identical, and understanding the differences determines whether the preview text earns its place in the email's first impression.
Preview Text Now Does Two Jobs: Human Inbox and AI Summary
In the pre-AI inbox era, preview text served one function: provide additional context to the human reader seeing the email in their inbox before deciding whether to open it. The preview text supplemented the subject line — the subject communicated the topic; the preview added specificity, urgency, or curiosity. The human either opened the email or did not. The preview text's job was done after that decision.
In the 2026 AI inbox era, this is still true — but a second function has been added. When Gemini AI or Apple Intelligence generates an inbox summary, it processes the subject line and the first text content of the email that it can parse. The preheader text — the text that appears as email preview in the inbox — is the first HTML text content in the email, and it is the first material the AI sees after the subject line. The AI's inbox summary is therefore substantially influenced by the preheader content. An informative preheader produces a useful AI summary; a vague or promotional preheader produces an AI summary that does not communicate the email's value.
The practical consequence: preview text that is written to tease or intrigue human readers ("You won't believe what we're announcing tomorrow") produces an AI summary that is similarly non-informative ("The email mentions an upcoming announcement"). Preview text that is specific and informative ("New: Summer linen collection in 8 new colours, available now for members") produces an AI summary that actually represents the email's content. The human opens the email based on the teaser; the AI cannot generate a useful summary from the teaser and the email earns lower inbox visibility in the relevance ranking.
How Gemini AI and Apple Intelligence Use Preview Text
Gmail Gemini AI (rolled out February 2026) generates a one-line inbox summary for emails when the user has AI features enabled. The summary appears in place of the standard preview text snippet in the inbox list view. The AI generates this summary by reading the email content — starting with the subject line and the first HTML text content it can parse. The preheader text, being the first HTML text in the email body, feeds directly into this generation process.
Apple Intelligence in Apple Mail (iOS 18/macOS Sequoia onwards) applies similar AI summarisation to email content. The Apple implementation tends to produce slightly longer summaries than Gemini's one-liner, and incorporates more of the email body. But the preheader text still provides the first textual context for Apple Intelligence's email content processing.
The AI systems do not distinguish between "preheader text" and "body text" — they read the HTML in document order and process the text they find. An email where the preheader is hidden via CSS (display:none or similar) presents no preheader text to the AI — the AI's summary starts from the first visible text in the email body, which may be a generic greeting, a navigation menu, or whatever comes first in the HTML structure. This is a meaningful difference from the human reading experience, where the hidden preheader appears in the inbox preview but not in the rendered email — the AI has no "inbox preview" layer and reads the actual HTML content.
Preheader Technical Implementation: The HTML That Works and the HTML That Doesn't
The preheader text appears in the email inbox preview by being the first text content in the HTML email body. Email clients read through the HTML and display the first text they find that has not been explicitly marked as hidden. The technical implementation requires balancing visibility in the inbox preview with hiding in the rendered email (most designs show the preheader in the preview but do not want a separate preheader text block visible in the opened email).
<!-- CORRECT preheader implementation (2026) --> <!-- Goes in <body> before the main email content --> <!-- Option 1: CSS hiding (most common, works well for human preview) --> <div style="display:none; max-height:0; overflow:hidden; mso-hide:all;"> Your preheader text goes here — specific, informative, under 90 characters. Padding characters to prevent the email client reading further into the email body: ‌ ‌ ‌ ‌ ‌ ‌ </div> <!-- This hides the preheader in the rendered email but shows in inbox preview --> <!-- PROBLEM FOR AI: Gemini and Apple Intelligence parse the full HTML --> <!-- display:none content IS read by AI systems in some implementations --> <!-- but is NOT reliable as the primary AI summary source --> <!-- Option 2: Visible preheader (better for AI, different visual design) --> <div style="color:#6b6b6b; font-size:12px; text-align:center; padding:8px 0;"> Summer linen collection — 8 new colours. <a href="[view-in-browser]">View in browser</a> </div> <!-- This is visible as a small text line at the top of the email --> <!-- AI systems read it reliably as the first text content --> <!-- Inbox preview shows it as expected --> <!-- Visual design: uses muted text that reads as a preheader rather than body copy --> <!-- Option 3: Hybrid (invisible preheader for human preview, visible summary for AI) --> <!-- Use invisible preheader for inbox preview as usual --> <div style="display:none; max-height:0; overflow:hidden;"> Short teaser for inbox preview — Summer linen in 8 new colours </div> <!-- Then place a visible but visually quiet text block that AI reads first --> <div style="color:#999; font-size:11px; text-align:center;"> New summer linen collection — 8 colours, member early access through Friday. </div> <!-- The invisible preheader controls what humans see in inbox preview --> <!-- The visible text block provides AI-readable content at the top of the email -->
The honest technical reality about AI preheader reading: different AI systems handle display:none content differently. Google's Gemini AI has been observed processing some display:none content in generating inbox summaries — meaning the hidden preheader text does feed into the AI summary in some cases. Apple Intelligence appears to be more selective about which hidden content it processes. The safest approach for reliable AI summary generation is to have meaningful, informative text in the visible email content, not to rely on the hidden preheader as the AI's primary source material.
Preview Text as a Spam Filter Signal
Email spam filters evaluate the first text content of an email as part of content scoring. The preheader text — being the first HTML text content — is weighted in this scoring. Several specific preheader patterns that generate negative spam filter signals:
"View this email in your browser" as the preview: This is the most common preheader failure mode — a link or text for viewing in browser that appears as the email's preview because no actual preheader text was set. From a spam filter's perspective: the email's first meaningful text is a navigation instruction with no substantive content — a pattern associated with spam that includes no readable body text. Additionally, from the human inbox experience: the "view in browser" preview communicates nothing about the email's content, significantly reducing open rates.
Promotional language heavy with spam trigger patterns: A preheader that reads "URGENT: Limited time offer — 80% OFF everything today only!!!" is accumulating spam filter scoring in the first line of the email. The subject line is already carrying some of this scoring load; the preheader adding a second layer of urgent promotional content compounds it. The preheader should complement the subject's tone rather than amplifying its promotional intensity.
Hidden preheader with colour-proximity text: As documented in the dark mode and HTML structure guides, setting preheader text with a colour nearly matching the background (white text on white background) to make it invisible triggers spam filter hidden-content detection. The correct approach for hidden preheaders is CSS display:none, not colour-proximity hiding.
Length and Content Strategy for 2026
Preview text length guidance has traditionally focused on inbox display limits — 85-100 characters is the common recommendation because that is approximately where most mobile inbox views truncate the preview. This guidance remains correct for human readability. For AI summarisation, the length consideration is different: AI systems read the full email content, not just the first 90 characters — so the preheader length does not limit what the AI can process. However, the first 40-50 characters of the preheader carry more weight in AI summary generation because they appear at the beginning of the text content, and AI language models weight earlier content more heavily when generating summaries.
The 2026 preview text content framework:
Characters 1-50: The specific, substantive claim. What this email is about, in concrete terms. "New summer linen in 8 colours" or "Your invoice for July is ready" or "Your order has shipped, arriving Thursday." This segment feeds AI summary generation and is what human readers see in truncated inbox previews on small screens.
Characters 51-90: Supporting context that adds value for human readers who see more preview text. Time-sensitive information ("member early access through Friday"), a specific benefit ("free shipping on orders over $75"), or an additional hook ("including 3 exclusive colourways"). This segment is truncated in many mobile previews but visible on desktop and in expanded inbox views.
Characters 91+: Padding characters to prevent the email client from displaying body text after the preheader. The standard padding is invisible characters: ‌ repeated. These prevent email clients from reaching into the email body for preview content but are invisible to humans and AI systems alike.
Testing Preview Text Across Clients and AI Systems
Testing preview text requires checking more surfaces than in previous years:
Standard inbox preview (human view): Send test emails to Gmail, Outlook.com, Apple Mail on iPhone, and Android Gmail. Check that the preheader appears correctly in the inbox preview, not "View this email in your browser" or HTML code or body text. Litmus and Email on Acid both render inbox preview thumbnails as part of their email client testing suites.
Gemini AI summary (Gmail): With a Gmail account that has Gemini AI features enabled (requires Gemini subscription or Workspace plan with AI features), send a test email and check the inbox summary line that Gemini generates. Compare the AI summary against the preheader text and the email's actual content — if the AI summary accurately represents the email's content, the preheader and opening content are working correctly. If the AI summary is vague or inaccurate, the preheader and opening content need revision.
Apple Intelligence summary (Apple Mail): With an iPhone running iOS 18 and Apple Intelligence enabled, send a test email to an iCloud address and check the inbox summary that Apple Intelligence generates. The Apple summary tends to be slightly longer and may incorporate more body content than Gemini's single-line summary.
This testing adds 15-20 minutes to the pre-send checklist but provides confirmation that the AI summary — which a significant fraction of the recipient list will see instead of the standard preview text — accurately represents the email's intent and content.
The Preview Text Mistakes That Hurt Deliverability
No preheader set (most common): The email renders to the inbox with "View this email in your browser" or the first HTML text from the navigation header as the preview. In Klaviyo, HubSpot, and most ESPs, the preheader is an optional field that defaults to nothing — meaning the email client reaches into the body for preview text, usually finding something useless. Fix: always set a preheader in the ESP before scheduling any campaign.
Preheader that repeats the subject line: "Summer Sale! → Don't miss our Summer Sale — up to 50% off" wastes the preview's opportunity to add information. The subject line and preview text should communicate different things. If the subject is the what, the preview should add the how, why, or when.
Preheader that is only padding characters: Some templates set a preheader of invisible characters only — ‌ repeated — to control what appears in the preview without setting actual text. The result: the inbox preview shows nothing, and the AI summary has no preheader content to start from. This is worse than not having a preheader at all, because it actively blocks both human preview and AI summarisation input.
Preheader with countdown timer or dynamic content failure: Some email platforms support dynamic preheader content (countdown timers, personalised variables). When these fail to render — because the recipient's email client does not support the dynamic element, or because the variable has no value — the preheader may display placeholder code ("[PREHEADER_TEXT]") or the raw countdown timer syntax. Test all dynamic preheader implementations with a null/failed state test to see what recipients see when the dynamic element does not work.
What High-Performing Preview Text Looks Like
High-performing preview text in 2026 is specific, complete as a standalone statement, and informative enough to generate a useful AI summary when the AI reads only the preview without the full email body. Examples across email types:
E-commerce promotional:
Subject: Summer sale starts now
Preview: 30% off linen, 40% off swimwear — free shipping orders over $75. Sale ends Sunday midnight.
The preview adds the discount specifics, the free shipping threshold, and the deadline — none of which are in the subject. An AI reading only this generates a useful summary: "Summer sale with 30% off linen, 40% off swimwear, free shipping over $75 through Sunday."
SaaS product update:
Subject: New in [Product]: Bulk export and custom fields
Preview: Export up to 50,000 records in CSV or JSON. Custom fields now support conditional logic.
Technical specifics that the subject cannot fit. AI summary: "Product update: bulk export up to 50,000 records and conditional logic for custom fields."
B2B newsletter:
Subject: This week in supply chain
Preview: Red Sea disruptions extend lead times 3-6 weeks; nearshoring adoption up 28% YoY; Q3 freight rate forecast.
Specific topic list that earns opens from readers with relevant interests. AI summary mirrors the preview accurately.
Transactional (order):
Subject: Your order #12847 has shipped
Preview: Arriving Thursday, April 24. 3 items via FedEx Ground. Tracking: 7489230498234.
Everything the recipient needs to know in the preview — they may not even need to open the email. For transactional email, this is the right design: make the information so accessible that the recipient's experience is positive regardless of whether they open or not.
The preview text is the most concise piece of email copy in each campaign — and in 2026, it is doing more work than ever. Treating it as an afterthought, or leaving it blank because "nobody really reads it," misses the compounding value of a preview text that works for human readers in the inbox, provides useful input to AI summarisation systems, and avoids the spam filter content signals that come from empty or generic preview content. Fifteen additional minutes per campaign to write a genuinely informative, specific preview text is the highest-ROI copy investment available in the email production process.