Page content, headings and image alt text

This chapter is about the words and structure on the page: the heading outline, pages with too little on them, pages that repeat each other, and images with no alt text. Six checks cover it, all listed on the audit checks page. The rule for all of them is the same: the assistant drafts, and you read everything before it goes live.

What the audit looks for

Two details help. The duplicate check compares the whole main text exactly, so pages that differ by one word, such as a service page copied for each suburb with only the suburb changed, will not be flagged. They are still near duplicates, so check for those by eye. And the thin check counts words in the main content, leaving out the header, navigation and footer.

Fix the heading outline

A page has one H1 that names its topic, then H2s for its sections, then H3s under those. Levels go in order, so no H4 straight after an H2. Page builders break this all the time, since a designer picked a heading level for its font size. The fix for that is to set the level correctly and style it to look the way the designer wanted.

Copy the outline from the page source, or from your audit's page detail. Give the assistant the levels along with the text.

Here is the heading outline of a page, copied from the page source (level and text):

{H1: ...
H2: ...
H3: ...}

The page is {what the page is for}. Return the outline fixed so that:
- there is exactly one H1, and it names the page's main topic
- levels go in order, with no skipped level
- each heading says what the section under it is about
Don't add or remove sections. List every change as: old heading, new heading, reason.

Thin pages: expand, merge or noindex

Do not reach for the assistant first. Decide what the page is for.

• The page answers something people search for (a service, a question). Expand it.
• The page overlaps another page. Merge it into the stronger one and redirect the old URL.
• The page has no reason to be found (a thank-you page, a login, a tag archive). Mark it noindex and stop worrying.

For the first case, the risk is the one from chapter 1: the assistant fills a short page with plausible facts the client never gave you. The prompt below forces a gap-marker in place of a guess, and gives you a list of questions for the client at the end. Those questions are the real value, because a page of 400 words that the client has confirmed beats a page of 1,000 that they have not.

This page has about {n} words and the audit marks it thin (under 200). The current text is:

{paste the text}

Draft an expanded version of 300 to 500 words that answers the questions a {customer type} asks before they {call / book / buy}. Use only facts in the client brief or the current text. Where a useful fact is missing (hours, process, area served, how long a job takes, a price range), write [CLIENT TO CONFIRM: what is needed] instead of guessing.

After the draft, list every [CLIENT TO CONFIRM] item as a short question I can send the client.

Google's guidance on helpful, people-first content is the standard to hold the draft to. Length is not on it.

Alt text for a batch of images

Alt text describes an image for a person who cannot see it, and helps search engines understand it. Write it for the person. A purely decorative image should keep an empty alt attribute, alt="", because that is correct for screen readers. Know one limit of the audit: it counts only non-empty alt text as covered, so a page full of decorative images can trip the check even when it is right. Judge those by eye. Export file names, the page each image is on, and the text next to it. An assistant that can read images does better if you attach them, but the text-only version is fine for most sites.

Write alt text for the images below. Rules:
- Describe what the image shows in 5 to 15 words, as you would to someone who can't see it.
- Mention the service or the place only when the image really shows it.
- Don't start with "image of" or "photo of".
- If an image looks purely decorative, answer alt="" and say why.

For each image I give you the file name, the page it is on and the text next to it:
{paste rows}

Return: file name | alt text.

Check a sample against the real images. File names like IMG_2041.jpg tell the assistant nothing, and it will invent something.

Don't bother with

• A word-count target. There is no number of words that ranks. The audit's 200 is a floor for "this page has almost nothing on it", not a goal.
• Stuffing the H1 with every phrase. One topic per page, and the H1 names it.
• Alt text full of keywords. It is for describing the image.
• Mass-producing pages. Google calls producing many pages with little value, with or without AI, scaled content abuse. Ten good pages beat a hundred thin ones.

Next

• Chapter 6: Structured data without hand-writing JSON-LD
• Track content fixes against each client
• Google: image SEO best practices (Google Search Central)