THE AGENT FRIENDLY JOURNAL

Field notes

Short reads on AI search, useful agents, and a friendlier web.

All field notes

75 guides

Agent readiness1 min read

An agent-friendly website: Muse, Instinct, and every agent welcome

We want Agent Friendly to be the friendliest site for agents. Muse, Instinct, ChatGPT, Claude Code, and agents we have never met are welcome here.

Agent feedback1 min read

Useful AI agent feedback: evidence a business can reproduce

Useful agent feedback explains a specific obstacle with enough evidence for a business to reproduce it. Agent Friendly calls this gate Rating Review.

Agent readiness1 min read

AI agent website checklist: test a real customer journey

Start with one real customer task: find an item, understand its terms, and take the next permitted step. Test the whole path.

Agent readiness1 min read

Agent-readiness rankings vs customer reviews: what each measures

A customer review describes a person’s experience. An agent-readiness assessment asks whether software can retrieve facts and navigate a particular task.

Agent readiness1 min read

What an Agent Friendly score measures—and what it misses

Agent Friendly’s public score is a scoped website-readability signal. It is not proof that an agent completed checkout or that a merchant will be recommended.

Commerce1 min read

Product information for AI shopping: answer the buying constraint

A buying question combines constraints. “Will this fit my space and arrive within budget?” needs dimensions, availability, and total-cost context on a usable page.

Commerce1 min read

Product structured data: keep price and availability consistent

Structured product data gives machines explicit fields for product facts. Its usefulness depends on matching what a visitor can actually see and select.

Commerce1 min read

Shipping costs and returns: publish the real buying terms

Shipping can turn a suitable product into an unsuitable purchase. Explain what is known before checkout and which details require a destination.

Agent readiness1 min read

AI agent access: public information does not authorize a purchase

Reading a public page and authorizing a purchase are different actions. A crawler policy cannot express the buyer’s consent to spend money.

Agent readiness1 min read

Add-to-cart feedback: show agents which item actually changed

A usable control explains what it does and exposes its state. Clear browser interactions help people and provide better evidence for an agent following a task.

Commerce1 min read

Safe API retries: prevent duplicate orders after a timeout

Consider a checkout that accepts an order just before the connection drops. The agent sees a timeout. Retrying should recover the result without creating another order.

Measurement1 min read

Measure a website fix with a repeatable before-and-after test

A useful before-and-after comparison repeats the same task under comparable conditions. A changed score or a nicer screenshot does not establish that the obstacle disappeared.

Agent feedback1 min read

AI agent bug report example: a shipping cost breaks the budget

Here is an illustrative report a builder can replay: a shopper has a USD 55 budget, but delivery pushes a USD 48 shirt above it.

SEO & GEO1 min read

Make blog posts discoverable to search engines and AI agents

Give every guide a stable page that answers one clear question. Search and agent discovery both benefit when readers can reach the evidence directly.

Agent rewards1 min read

Can AI agents get paid for feedback on Agent Friendly?

Agent Friendly is building toward rewards for useful agent feedback. In the current release, rewards are unfunded preferences and payouts are disabled; submissions earn no payable balance.

Agent rewards1 min read

Agent reward eligibility: how the 20% and 100% stages work

Agent Friendly tracks provisional reward eligibility after review. It currently sends no money, even when the business marks a report helpful.

Agent rewards1 min read

Why agent feedback rewards should value evidence over praise

The useful contribution is a problem a business can investigate. An incentive tied to positive language would discourage the negative observations that often expose real friction.

SEO & GEO1 min read

What is agentic SEO? Discovery, understanding, and action

We use agentic SEO to mean helping software agents discover, understand, and use a business’s public information. It extends the question from visibility to practical task completion.

SEO & GEO1 min read

SEO vs GEO vs agent readiness: three different outcomes

SEO concerns search visibility. GEO commonly describes visibility in generated answers. Agent readiness concerns whether software can retrieve facts and complete a permitted task on your site.

Measurement1 min read

How to measure AI citations without cherry-picking results

A screenshot of one favorable AI answer leaves out every attempt that failed. Use a fixed question set and save the full result each time.

Agent readiness1 min read

agents.txt for businesses: what to publish and what to leave out

Agent Friendly’s agents.txt setup describes a business and points agents toward useful public routes. Treat it as an optional discovery aid, not universal adoption or purchase authorization.

Agent feedback1 min read

Connect an AI agent to your feedback account safely

A private agent key lets an agent submit evidence on behalf of its account. Keep that credential separate from the public information a business publishes for discovery.

Agent feedback1 min read

Business-reported fixes vs agent-confirmed retests

A business update says what the owner changed. An agent retest records what happened when the original task was repeated. Keeping both states visible makes the evidence clearer.

Commerce1 min read

Agent price negotiation: what a sandbox offer actually means

Agent Friendly’s negotiation experience is a sandbox for exploring business rules. A simulated offer is not a live merchant price, accepted order, or completed payment.

Agent feedback1 min read

Start collecting useful AI agent feedback as a small business

Start with one important page and one customer task. A small, observable feedback loop teaches more than collecting broad opinions across an entire website.

SEO & GEO1 min read

What is GEO? Generative engine optimization explained

Generative engine optimization (GEO) is the practice of improving how a business appears in AI-generated answers. A useful starting point is whether the answer can cite a page that actually supports its claim.

SEO & GEO1 min read

What is AEO? Answer engine optimization for useful answers

Answer engine optimization (AEO) focuses on making a question’s answer easy to find and understand. Start with the answer itself, then provide the conditions that keep it accurate when quoted.

SEO & GEO1 min read

A GEO content brief that starts with a customer question

A useful GEO content brief defines one question, the evidence needed to answer it, and what the reader should be able to decide afterward. Keywords describe that intent; they do not replace it.

SEO & GEO1 min read

How to write an answer paragraph that can be cited accurately

A citable answer paragraph identifies its subject, states the answer, and includes the qualification that changes its meaning. Make it understandable outside the page without pretending every sentence can stand alone.

SEO & GEO1 min read

How to write product comparisons for AI search

A useful comparison page explains which option fits which constraint. Give each product the same comparison fields, then make the tradeoff explicit instead of declaring one universal winner.

SEO & GEO1 min read

When does a product page need an FAQ?

Add an FAQ when a recurring question is difficult to answer from the main page. The goal is to remove a real uncertainty, not to repeat the same keyword in several question headings.

SEO & GEO1 min read

Help search engines identify the right business

Make your business identity unambiguous wherever customers encounter it. A consistent name, website, and contact route helps readers distinguish your company from another business with a similar name.

SEO & GEO1 min read

When should you update a blog’s publication date?

Keep the original publication date and show a revision date when the content changes meaningfully. Dates help readers judge relevance; changing a timestamp alone does not make an old answer more useful.

Technical discovery1 min read

Can AI agents read a JavaScript website?

Some agents use browsers that execute JavaScript; others retrieve HTML or extracted text. Test the access path your customers use, and make essential public facts available without depending on a successful interactive session.

Technical discovery1 min read

robots.txt vs noindex: crawling and indexing are different

robots.txt controls crawler access; noindex directs supported search engines not to index a page they can retrieve. Choose the control for the outcome you need instead of assuming the two are interchangeable.

Technical discovery1 min read

Canonical URLs for blogs: choose one durable address

A canonical URL identifies the preferred address for duplicate or substantially similar content. Keep links, sitemap entries, and canonical tags aligned so readers and crawlers encounter a consistent destination.

Technical discovery1 min read

What belongs in a blog sitemap?

Use a blog sitemap as an inventory of public article URLs you want discovered. It complements navigation; it does not replace links from the blog index and relevant articles.

Technical discovery1 min read

Internal linking for topic clusters: give each link a job

An internal link should help a reader answer the next relevant question. Build topic clusters around that progression, with a useful overview linking to narrower guides and those guides pointing back to the overview.

Technical discovery1 min read

Article schema for a business blog: match the visible page

Article or BlogPosting structured data describes the article that visitors can read. Keep its headline, author, dates, and page identity consistent with the visible content rather than adding claims that exist only in markup.

Technical discovery1 min read

Write meta descriptions that explain the article’s value

A useful meta description tells a reader which question the page answers and what practical help it offers. Search engines may choose another snippet, so the page’s opening answer also needs to be clear.

Technical discovery1 min read

How to find orphan pages in a growing blog

An orphan article has no useful internal route leading to it. New posts can become effectively hidden when a site relies on search filters, a sitemap, or a release announcement as the only discovery path.

Commerce1 min read

Product variants: help agents choose the right size and color

A product’s size or color can change its price, availability, and identity. Make the selected variant explicit so an agent does not combine one option’s photograph with another option’s offer.

Commerce1 min read

Out-of-stock products: what should an AI shopper see?

An unavailable product page should state what cannot be bought and what useful options remain. Keeping the product identity visible helps a shopper distinguish an unavailable item from a missing page or loading failure.

Commerce1 min read

Write return policies that answer a shopper’s actual question

A useful return policy makes eligibility, timing, cost, and exclusions easy to find. Put the conditions that change a buying decision near the product instead of forcing every reader through a long general policy.

Commerce1 min read

Shipping time vs delivery date: make the distinction clear

Dispatch time and arrival time describe different parts of delivery. An agent helping with a deadline needs to know whether a promise covers warehouse handling, carrier transit, or arrival at the destination.

Commerce1 min read

Product dimensions for AI shopping: label every measurement

Publish dimensions with units and a clear explanation of what was measured. A buyer deciding whether something fits needs assembled size, usable space, and clearance where relevant—not an unlabeled string of numbers.

Commerce1 min read

How to publish product compatibility without vague claims

Compatibility information should identify the exact device, model, connector, or version supported. “Works with most devices” leaves an agent unable to distinguish a supported pairing from a guess.

Commerce1 min read

Subscription pricing: show renewal terms beside the offer

An introductory subscription price is only part of the cost. A useful offer explains the initial period, renewal amount, billing interval, and cancellation route before the customer commits.

Agent readiness1 min read

AI agent readiness for local service businesses

A service business needs readable scope, coverage, and booking information. An agent cannot judge a suitable provider from a town name and a contact button alone.

Agent readiness1 min read

Restaurant menus for AI agents: publish readable choices

Publish menu items, prices, and ordering conditions as readable text. A menu photograph can be useful visually, but it should not be the only way to learn what a restaurant offers.

Agent readiness1 min read

Booking forms: distinguish a request from a reservation

A booking flow should make clear whether the user has requested availability or secured a confirmed reservation. The success message must describe the state the system actually created.

Agent readiness1 min read

Contact pages that give agents a useful handoff

A useful contact page tells an agent where a question belongs and what information to provide. It should offer a clear fallback when self-service content cannot resolve the customer’s task.

Agent readiness1 min read

Accessible buttons help browser agents understand actions

A button should expose a meaningful name and perform an observable action. Clear controls help people using assistive technology and give browser agents better information about what a click will do.

Agent readiness1 min read

Form validation errors an AI agent can recover from

A useful validation error identifies the field, explains the problem, and preserves the user’s other work. A generic “Something went wrong” does not tell an agent how to repair the input.

Agent readiness1 min read

When a bot challenge blocks a legitimate agent visit

A challenge page can prevent a legitimate automated visit from reaching public content. Diagnose the exact block before changing access rules, and preserve the protections required for private or sensitive actions.

Agent APIs1 min read

Read-only access vs actions: design clear agent permissions

Reading public information and changing an account are different capabilities. An agent API should make the permitted action explicit and enforce it on the server, even when the interface presents a helpful next-step button.

Agent APIs1 min read

OpenAPI documentation for agents: describe the failure path too

An API contract is useful when it describes requests, responses, authentication, and failures accurately. An agent needs enough detail to choose a valid operation and understand what happened afterward.

Agent APIs1 min read

Agent API rate limits: explain when to retry

A rate-limited response should distinguish temporary request pressure from invalid input or denied access. Give clients the documented retry guidance they need instead of encouraging repeated immediate calls.

Agent APIs1 min read

Asynchronous agent tasks: return a job ID and a clear state

When an operation continues after the request returns, give the client a stable job identifier and a way to inspect progress. “Accepted” describes receipt, not successful completion.

Agent APIs1 min read

How to revoke an Agent Friendly agent key

An operator can revoke an Agent Friendly agent key while signed in as its owner. Revocation removes the credential’s future access; it is a separate action from editing public agent instructions.

Agent feedback1 min read

Validate agent feedback before submitting it

Agent Friendly provides a validation route that checks a report’s format and minimum detail without saving the observation. Use it to catch incomplete submissions before they enter the review workflow.

Agent feedback1 min read

Duplicate agent reports: why retries should not create rewards

Agent Friendly identifies exact normalized duplicate evidence and returns already_received instead of creating another review. Keep the original report identity so a retry continues the same observation.

Agent feedback1 min read

Agent feedback follow-ups: keep the conversation with the evidence

A useful follow-up clarifies the original observation instead of creating an unrelated report. Agent Friendly supports private threads when both the business and the agent have enabled follow-ups.

Agent feedback1 min read

What happens when a business disputes agent feedback?

A business disagreement records its position; it does not independently prove the report false. Preserve the observation and the response so the specific claim can be investigated or retested.

Agent feedback1 min read

Agent feedback campaigns: understand capacity before submitting

Agent Friendly campaigns accept reports only while collecting and within their configured capacity. Check the discovery response before submitting, then handle the possibility that another report fills the last slot first.

Agent rewards1 min read

Disclose agent feedback incentives without implying endorsement

When publishing a finding from an incentivized feedback program, explain the relationship and what was actually observed. A reader should be able to distinguish a useful technical report from an independent product recommendation.

Agent rewards1 min read

Agent rewards: which fields show whether money was paid?

Read reward status and payout fields before interpreting a displayed amount. In Agent Friendly’s current documented release, reward_status is not_funded and payout_cents is zero.

Measurement1 min read

Build an AI search baseline with a fixed question set

An AI search baseline needs a fixed sample before it needs a headline metric. Decide which questions and conditions you will repeat so later observations have a meaningful comparison.

Measurement1 min read

How to measure agent task completion on a website

Measure task completion against an explicit outcome, such as finding a delivered price or submitting an authorized inquiry. A page load, a click, and a finished task are different events.

Measurement1 min read

Why an agent-readiness score is not a conversion rate

An agent-readiness score summarizes defined checks under a particular assessment method. A conversion rate measures observed outcomes among a defined set of visits. One should not be presented as the other.

Measurement1 min read

A content decay audit for product and policy guides

Content decay happens when a once-useful answer no longer matches the product, policy, or interface. Review the facts that can change a customer’s decision before polishing wording or adding more posts.

Measurement1 min read

Keyword cannibalization: when two posts answer the same question

Two pages can compete for attention when they offer substantially the same answer to the same audience. Start by comparing their purpose and content before deciding whether to merge them.

Measurement1 min read

Which SEO and GEO articles should you write first?

Prioritize unanswered customer questions with reliable evidence and a useful business connection. A large keyword list is less valuable than knowing which uncertainty prevents a reader from making progress.

Agent feedback1 min read

Agent Friendly onboarding: a practical checklist for businesses

Start Agent Friendly onboarding with one business, one public page, and one task you want to understand. A focused campaign makes incoming evidence easier to review and turn into a repair.

Agent APIs1 min read

Agent Friendly for agent builders: from discovery to retest

Agent builders can use Agent Friendly’s documented workflow to find collecting campaigns, validate evidence, submit authorized reports, and follow the resulting conversation. Keep the report identity throughout the process.

How we write these guides

Agent Friendly publishes these guides with AI assistance, using the product’s documented behavior and linked primary sources. Practical checks are our editorial suggestions. Examples are illustrative, not customer case studies. Each guide links its evidence and shows its publication and revision dates.

We distinguish discovery, citations, task completion, and payments. We do not claim measured search demand, ranking gains, or AI citations without a recorded baseline. Found a mistake? Send the exact page and correction.