Lesson 8 · 18 min
Agents already fill in forms and compare products by reading the page the way a screen reader does. What helps is the basics done well: semantic HTML, labels, correct product data. And none of it counts without a routine: a monthly loop with the dashboards, the battery, the log and the score, always with a date.
Start the lesson
Goal: make forms and product pages readable for agents and set up the monthly loop for measuring in the AIs, with dates.
Click each block.
The basics made readable for agents, plus a routine with a date, are worth more than any shortcut.
Click each step. All of them come from web.dev's guide to agent-friendly sites.
An action is a button, navigation is a link. If you can't use the right element, give it the correct role and tabindex.
Use a hand cursor (cursor: pointer) on what can be clicked: it's a strong signal of an action.
Every field has a visible label tied to it with the for attribute. Email, name, and phone with autocomplete: it helps the person filling in the form and helps the agent know what each field is.
No transparent layers over buttons; action elements visible and at a minimum size. A layout that doesn't move around after loading: an agent that reads a screenshot gets lost.
A correct Merchant Center feed, a clear return policy, and an up-to-date Business Profile. Follow agent commerce protocols without implementing them: sources still disagree about their status.
The form is a template; the loop sheet is for copying every month.
<form action="/orcamento" method="post"> <label for="nome">Name</label> <input id="nome" name="nome" autocomplete="name" required> <label for="email">Email</label> <input id="email" name="email" type="email" autocomplete="email" required> <label for="telefone">Phone</label> <input id="telefone" name="telefone" type="tel" autocomplete="tel"> <button type="submit">Request a quote</button> </form>
# Monthly AI loop — <month/year> (run on dd/mm/yyyy) 1. Dashboards: GSC AI impressions <n> | Bing citations <n> | cited pages <n> 2. Battery v1: appearance in the neutral questions by assistant: ChatGPT <x%> | Gemini <x%> | Perplexity <x%> | Copilot <x%> Sources cited in my place (top 5): <...> 3. Brand: "what is [brand]" correct in <n of 3> assistants; "who is [founder]" current in <n of 3> 4. Log: search <n> | user action <n> | training <n> | false <n> | 403 for AI bot: <yes/no> 5. Readiness score: <n> (tool and version) Decision of the month (just one): <...> No change for 3 months? External sources to work on: <...>
The AIV method suggests: after 3 months with no change in the battery, the lever is no longer your site's content but your presence where the AIs already look (the sources that show up in your place). Treat it as a hypothesis to test, with the loop measuring.
E2 uses the generator at the end of the lesson.
Pick the most important form on your site (contact, quote, order). In Chrome DevTools, check the accessibility tree: does each field have a name? Is the button a button? Fix whatever is missing.
Done when: every form field appears by name in the accessibility tree, email, name, and phone have autocomplete, and submitting is a button.
Generate your plan below, download the .ics, and put the first monthly loop on your calendar. Copy the loop sheet into the project folder.
Done when: the calendar has the date of the first monthly loop, and the month's loop-ia sheet exists with the date filled in.
The explanation appears after your answer.
Question 1 of 4
What does a browser agent use to understand a form?skill
Question 2 of 4
What should a small business do about agentic commerce protocols today?
Question 3 of 4
The appearance rate in the battery has been flat for 3 months. Next step?
Question 4 of 4
Why does the loop have a date and one decision per month?
Choose an answer. The explanation appears after you try.
Four steps and the exercises for each. Click an exercise to see its "done when".
foundation × fad
2 exercises
Triage with sources and the passages only you have.
View the lessondashboards and battery (lessons 2 and 3)
2 exercises
Baseline in the dashboards and the first round of the battery.
View the lessonentity, log, and grade (lessons 4 to 6)
2 exercises
Identity fixed, bots checked, grade with one fix.
View the lessonagents and the loop (lessons 7 and 8)
2 exercises
WebMCP test page, readable form, scheduled loop.
View the lessonBefore measuring, know what is worth doing.
Requires the triage. The map shows the exercises from lesson 3; the plan below lists them all.
Requires the measurement: fixing without a baseline shows no effect.
Requires the rest. From here on, the monthly loop measures everything.
Separate, measure, fix, and prepare. Only move up a step when the "done when" of the previous one is true.
After this course comes number 5: agents in the terminal doing the repetitive work, with gates.
Pick your start date and how much time you have per week. The plan respects the exercises that require waiting days.